{ "timestamp": "2026-09-21T08:02:26.830059", "event": "schedule", "run_identity": { "attempt_id": "", "repository_commit": "19694d9d9ba7cef51bce0e5ee7229041802bcb9e" }, "selection": { "mode": "full", "expected_executions": [ { "kind": "python", "id": "gemm", "args": [] }, { "kind": "python", "id": "gemm_variants", "args": [] }, { "kind": "python", "id": "gemm_microkernel", "args": [] }, { "kind": "python", "id": "gemm_fused", "args": [] }, { "kind": "python", "id": "relu", "args": [] }, { "kind": "python", "id": "gelu", "args": [] }, { "kind": "python", "id": "sigmoid", "args": [] }, { "kind": "python", "id": "softmax", "args": [] }, { "kind": "python", "id": "swiglu", "args": [] }, { "kind": "python", "id": "layernorm", "args": [] }, { "kind": "python", "id": "rmsnorm", "args": [] }, { "kind": "python", "id": "rope", "args": [] }, { "kind": "python", "id": "embedding", "args": [] }, { "kind": "python", "id": "attention", "args": [] }, { "kind": "python", "id": "attention_sliding", "args": [] }, { "kind": "python", "id": "kv_cache", "args": [] }, { "kind": "python", "id": "kv_cache_decode", "args": [] }, { "kind": "python", "id": "fused_attention_decode", "args": [] }, { "kind": "python", "id": "fused_swiglu_decode", "args": [] }, { "kind": "python", "id": "mlp", "args": [] }, { "kind": "python", "id": "cross_entropy", "args": [] }, { "kind": "python", "id": "optimizer", "args": [] }, { "kind": "python", "id": "lm_head_litmus", "args": [] }, { "kind": "python", "id": "deepseek_reference", "args": [] }, { "kind": "python", "id": "yarn_rope_explicit_positions", "args": [] }, { "kind": "python", "id": "relu2", "args": [] }, { "kind": "python", "id": "nemotron_router", "args": [] }, { "kind": "python", "id": "moe_relu2_expert", "args": [] }, { "kind": "python", "id": "mamba2_reference", "args": [] }, { "kind": "python", "id": "gemma4_assistant", "args": [] }, { "kind": "python", "id": "vision", "args": [] }, { "kind": "python", "id": "audio_frontend", "args": [] }, { "kind": "python", "id": "audio_transformer_primitives", "args": [] }, { "kind": "python", "id": "nightly_runner_hardware", "args": [] }, { "kind": "python", "id": "idle_nightly_coordinator", "args": [] }, { "kind": "python", "id": "relu_bf16", "args": [] }, { "kind": "python", "id": "gelu_bf16", "args": [] }, { "kind": "python", "id": "sigmoid_bf16", "args": [] }, { "kind": "python", "id": "swiglu_bf16", "args": [] }, { "kind": "python", "id": "layernorm_bf16", "args": [] }, { "kind": "python", "id": "rmsnorm_bf16", "args": [] }, { "kind": "python", "id": "rope_bf16", "args": [] }, { "kind": "python", "id": "yarn_rope_explicit_positions_bf16", "args": [] }, { "kind": "python", "id": "embedding_bf16", "args": [] }, { "kind": "python", "id": "attention_bf16", "args": [] }, { "kind": "python", "id": "mlp_bf16", "args": [] }, { "kind": "python", "id": "cross_entropy_bf16", "args": [] }, { "kind": "python", "id": "vision_position_storage_bf16", "args": [] }, { "kind": "python", "id": "v8_bf16_safetensors_lowering", "args": [] }, { "kind": "python", "id": "v8_bf16_file_backed_bump", "args": [] }, { "kind": "python", "id": "v8_instella_moe_circuit_contracts", "args": [] }, { "kind": "python", "id": "pytorch_onednn_oracle_contract_bf16", "args": [] }, { "kind": "python", "id": "q4k_kernels", "args": [] }, { "kind": "python", "id": "q6k_kernels", "args": [] }, { "kind": "python", "id": "q4_k_q8_k_matvec", "args": [] }, { "kind": "python", "id": "softmax_backward", "args": [] }, { "kind": "python", "id": "attention_backward", "args": [] }, { "kind": "python", "id": "deltanet_backward", "args": [] }, { "kind": "python", "id": "pytorch_parity", "args": [] }, { "kind": "python", "id": "qwen3vl_private_corpus_contract", "args": [] }, { "kind": "python", "id": "v8_generated_layout_bridges", "args": [] }, { "kind": "python", "id": "rope_pairwise_layout", "args": [] }, { "kind": "make", "id": "test-native-session-v8", "args": [] }, { "kind": "make", "id": "test-server-schema", "args": [] }, { "kind": "make", "id": "test-audio-v8-contracts", "args": [] }, { "kind": "make", "id": "test-tokenizer-special", "args": [] }, { "kind": "make", "id": "litmus", "args": [] }, { "kind": "make", "id": "llamacpp-parity-nightly", "args": [] }, { "kind": "make", "id": "test-flash-attention", "args": [] }, { "kind": "make", "id": "test-attention-f16-split-kv", "args": [] }, { "kind": "make", "id": "test-q4q6-llama-production-native", "args": [] }, { "kind": "make", "id": "test-q4k-q8k-llama-performance", "args": [] }, { "kind": "make", "id": "test-qwen36-q6k-m4-performance", "args": [] }, { "kind": "make", "id": "test-whisper-f16-gemm-performance", "args": [] }, { "kind": "make", "id": "test-private-numerical-replay-runner", "args": [] }, { "kind": "make", "id": "test-v8-dump-alignment", "args": [] }, { "kind": "make", "id": "test-threadpool-parity", "args": [] }, { "kind": "make", "id": "v7-kernel-map-contracts", "args": [] }, { "kind": "make", "id": "regression-training-full", "args": [] }, { "kind": "make", "id": "v8-training-certify-fp32", "args": [] }, { "kind": "make", "id": "v8-training-workflow-fp32", "args": [] }, { "kind": "make", "id": "v8-training-matrix-nightly", "args": [] }, { "kind": "make", "id": "v7-backprop-long-epoch-nightly", "args": [] }, { "kind": "make", "id": "v7-kernel-parity-train", "args": [] }, { "kind": "make", "id": "v7-ir-visualizer-e2e-nightly", "args": [] }, { "kind": "make", "id": "v7-visualizer-health", "args": [] }, { "kind": "make", "id": "v7-visualizer-generated-e2e", "args": [] }, { "kind": "make", "id": "v7-stabilization-nightly", "args": [] }, { "kind": "make", "id": "v8-validate-contracts", "args": [] }, { "kind": "make", "id": "v8-kernel-map-contracts", "args": [] }, { "kind": "make", "id": "v8-model-memory-plans-nightly", "args": [] }, { "kind": "make", "id": "test-numerical-contracts", "args": [] }, { "kind": "make", "id": "test-v8-cohere-laguna-contracts", "args": [] }, { "kind": "make", "id": "test-v8-artifact-compile-matrix", "args": [] }, { "kind": "make", "id": "test-v8-muse-glimmer-contracts", "args": [] }, { "kind": "make", "id": "test-v8-capability-cases", "args": [] }, { "kind": "make", "id": "test-v8-qwen38-dense-contracts", "args": [] }, { "kind": "make", "id": "test-v8-routed-moe-contracts", "args": [] }, { "kind": "make", "id": "test-v8-qwen38-flash-contracts", "args": [] }, { "kind": "make", "id": "test-v8-command-a-plus-nvfp4", "args": [] }, { "kind": "make", "id": "test-bf16-xray", "args": [] }, { "kind": "make", "id": "test-xray-validator-selftest", "args": [] }, { "kind": "make", "id": "test-bf16-pytorch-onednn-oracle-auto", "args": [] }, { "kind": "make", "id": "test-qwen3vl-bf16-kernel-oracles", "args": [] }, { "kind": "make", "id": "test-qwen3vl-bf16-private-corpus-parity-auto", "args": [] }, { "kind": "make", "id": "test-v8-dsl-policy", "args": [] }, { "kind": "make", "id": "test-source-integrity", "args": [] }, { "kind": "make", "id": "test-v8-template-circuit-audit", "args": [] }, { "kind": "make", "id": "test-architecture-contracts", "args": [] }, { "kind": "make", "id": "v8-regression-fast", "args": [] }, { "kind": "make", "id": "v8-visualizer-vision-artifacts", "args": [] }, { "kind": "make", "id": "test-v8-vision-kernels", "args": [] }, { "kind": "make", "id": "test-v8-qwen3vl", "args": [] }, { "kind": "make", "id": "test-qwen3vl-methodical-parity", "args": [] }, { "kind": "make", "id": "test-qwen3vl-private-corpus-parity-auto", "args": [] }, { "kind": "make", "id": "test-qwen36vl-private-corpus-parity-auto", "args": [] }, { "kind": "make", "id": "test-gemma4-private-corpus-parity-auto", "args": [] }, { "kind": "make", "id": "test-v8-qwen3vl-e2e-smoke", "args": [] }, { "kind": "make", "id": "vision-encoder-full", "args": [] }, { "kind": "make", "id": "test-v8-gemma4-vision-smoke", "args": [] }, { "kind": "make", "id": "test-v8-decoder-matrix-quick", "args": [] }, { "kind": "make", "id": "test-v8-gemma4-highmem", "args": [] }, { "kind": "make", "id": "test-v8-nemotron9-highmem", "args": [] }, { "kind": "make", "id": "test-v8-glm4-highmem", "args": [] }, { "kind": "make", "id": "test-v8-kimi-highmem", "args": [] }, { "kind": "make", "id": "test-v8-qwen36-highmem", "args": [] }, { "kind": "make", "id": "test-v8-xeon-family-contracts", "args": [] }, { "kind": "make", "id": "test-v8-serve-localhost-e2e", "args": [] }, { "kind": "make", "id": "bench_gemm", "args": [] }, { "kind": "make", "id": "bench-q4k-dispatch-matrix-quick", "args": [] }, { "kind": "make", "id": "bench-q8-0-fp32-gemm-quick", "args": [] }, { "kind": "make", "id": "bench-q4k-gateup-swiglu-x16-chunk4-quick", "args": [] } ] }, "duration_sec": 3176.1170570850372, "runner_python": { "executable": "/opt/hostedtoolcache/Python/3.11.16/x64/bin/python", "version": "3.11.16" }, "runner_hardware": { "schema": 1, "captured_at": "2026-09-21T08:55:23.158199Z", "available": true, "provider": "github-actions", "cpu_source": "lscpu", "runner": { "name": "GitHub Actions 1000006097", "os": "Linux", "arch": "X64", "image_os": "ubuntu24", "image_version": "20260907.300.1", "kernel": "6.17.0-1022-azure", "workflow": "Nightly Tests", "run_id": "35574289723", "run_attempt": "1" }, "cpu": { "architecture": "x86_64", "vendor_id": "AuthenticAMD", "model_name": "AMD EPYC 7763 64-Core Processor", "family": "25", "model": "1", "stepping": "1", "logical_cpus": "4", "online_cpus": "0-3", "sockets": "1", "cores_per_socket": "2", "threads_per_core": "2", "numa_nodes": "1", "max_mhz": "", "min_mhz": "", "virtualization": "AMD-V" }, "isa": { "avx": true, "avx2": true, "fma": true, "avx_vnni": false, "avx512f": false, "avx512_vnni": false, "avx512_bf16": false, "amx_tile": false, "amx_int8": false, "amx_bf16": false }, "memory": { "total": "16373448 kB", "available": "15169204 kB", "free": "4554508 kB", "swap_total": "3145724 kB", "swap_free": "3145684 kB" }, "lscpu_fields": { "Architecture": "x86_64", "CPU op-mode(s)": "32-bit, 64-bit", "Address sizes": "48 bits physical, 48 bits virtual", "Byte Order": "Little Endian", "CPU(s)": "4", "On-line CPU(s) list": "0-3", "Vendor ID": "AuthenticAMD", "Model name": "AMD EPYC 7763 64-Core Processor", "CPU family": "25", "Model": "1", "Thread(s) per core": "2", "Core(s) per socket": "2", "Socket(s)": "1", "Stepping": "1", "BogoMIPS": "4890.86", "Flags": "fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl tsc_reliable nonstop_tsc cpuid extd_apicid aperfmperf tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm cmp_legacy svm cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw topoext vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves user_shstk clzero xsaveerptr rdpru arat npt nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload umip vaes vpclmulqdq rdpid fsrm", "Virtualization": "AMD-V", "Hypervisor vendor": "Microsoft", "Virtualization type": "full", "L1d cache": "64 KiB (2 instances)", "L1i cache": "64 KiB (2 instances)", "L2 cache": "1 MiB (2 instances)", "L3 cache": "32 MiB (1 instance)", "NUMA node(s)": "1", "NUMA node0 CPU(s)": "0-3", "Vulnerability Gather data sampling": "Not affected", "Vulnerability Ghostwrite": "Not affected", "Vulnerability Indirect target selection": "Not affected", "Vulnerability Itlb multihit": "Not affected", "Vulnerability L1tf": "Not affected", "Vulnerability Mds": "Not affected", "Vulnerability Meltdown": "Not affected", "Vulnerability Mmio stale data": "Not affected", "Vulnerability Old microcode": "Not affected", "Vulnerability Reg file data sampling": "Not affected", "Vulnerability Retbleed": "Not affected", "Vulnerability Spec rstack overflow": "Vulnerable: Safe RET, no microcode", "Vulnerability Spec store bypass": "Vulnerable", "Vulnerability Spectre v1": "Mitigation; usercopy/swapgs barriers and __user pointer sanitization", "Vulnerability Spectre v2": "Mitigation; Retpolines; STIBP disabled; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected", "Vulnerability Srbds": "Not affected", "Vulnerability Tsa": "Vulnerable: No microcode", "Vulnerability Tsx async abort": "Not affected", "Vulnerability Vmscape": "Not affected" }, "raw_lscpu": "Architecture: x86_64\nCPU op-mode(s): 32-bit, 64-bit\nAddress sizes: 48 bits physical, 48 bits virtual\nByte Order: Little Endian\nCPU(s): 4\nOn-line CPU(s) list: 0-3\nVendor ID: AuthenticAMD\nModel name: AMD EPYC 7763 64-Core Processor\nCPU family: 25\nModel: 1\nThread(s) per core: 2\nCore(s) per socket: 2\nSocket(s): 1\nStepping: 1\nBogoMIPS: 4890.86\nFlags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl tsc_reliable nonstop_tsc cpuid extd_apicid aperfmperf tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm cmp_legacy svm cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw topoext vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves user_shstk clzero xsaveerptr rdpru arat npt nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload umip vaes vpclmulqdq rdpid fsrm\nVirtualization: AMD-V\nHypervisor vendor: Microsoft\nVirtualization type: full\nL1d cache: 64 KiB (2 instances)\nL1i cache: 64 KiB (2 instances)\nL2 cache: 1 MiB (2 instances)\nL3 cache: 32 MiB (1 instance)\nNUMA node(s): 1\nNUMA node0 CPU(s): 0-3\nVulnerability Gather data sampling: Not affected\nVulnerability Ghostwrite: Not affected\nVulnerability Indirect target selection: Not affected\nVulnerability Itlb multihit: Not affected\nVulnerability L1tf: Not affected\nVulnerability Mds: Not affected\nVulnerability Meltdown: Not affected\nVulnerability Mmio stale data: Not affected\nVulnerability Old microcode: Not affected\nVulnerability Reg file data sampling: Not affected\nVulnerability Retbleed: Not affected\nVulnerability Spec rstack overflow: Vulnerable: Safe RET, no microcode\nVulnerability Spec store bypass: Vulnerable\nVulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2: Mitigation; Retpolines; STIBP disabled; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected\nVulnerability Srbds: Not affected\nVulnerability Tsa: Vulnerable: No microcode\nVulnerability Tsx async abort: Not affected\nVulnerability Vmscape: Not affected", "raw_cpuinfo": "", "raw_free_h": "total used free shared buff/cache available\nMem: 15Gi 1.1Gi 4.3Gi 52Mi 10Gi 14Gi\nSwap: 3.0Gi 40Ki 3.0Gi", "raw_meminfo": "MemTotal: 16373448 kB\nMemFree: 4554508 kB\nMemAvailable: 15169204 kB\nBuffers: 490884 kB\nCached: 10072560 kB\nSwapCached: 40 kB\nActive: 1008496 kB\nInactive: 9888264 kB\nActive(anon): 332484 kB\nInactive(anon): 63104 kB\nActive(file): 676012 kB\nInactive(file): 9825160 kB\nUnevictable: 46776 kB\nMlocked: 43704 kB\nSwapTotal: 3145724 kB\nSwapFree: 3145684 kB\nZswap: 0 kB\nZswapped: 0 kB\nDirty: 4032 kB\nWriteback: 0 kB\nAnonPages: 365296 kB\nMapped: 314732 kB\nShmem: 53524 kB\nKReclaimable: 459476 kB\nSlab: 591308 kB\nSReclaimable: 459476 kB\nSUnreclaim: 131832 kB\nKernelStack: 4032 kB\nPageTables: 6692 kB\nSecPageTables: 0 kB\nNFS_Unstable: 0 kB\nBounce: 0 kB\nWritebackTmp: 0 kB\nCommitLimit: 11332448 kB\nCommitted_AS: 2102832 kB\nVmallocTotal: 34359738367 kB\nVmallocUsed: 39344 kB\nVmallocChunk: 0 kB\nPercpu: 2256 kB\nHardwareCorrupted: 0 kB\nAnonHugePages: 180224 kB\nShmemHugePages: 0 kB\nShmemPmdMapped: 0 kB\nFileHugePages: 1316864 kB\nFilePmdMapped: 0 kB\nCmaTotal: 0 kB\nCmaFree: 0 kB\nUnaccepted: 0 kB\nBalloon: 0 kB\nHugePages_Total: 0\nHugePages_Free: 0\nHugePages_Rsvd: 0\nHugePages_Surp: 0\nHugepagesize: 2048 kB\nHugetlb: 0 kB\nDirectMap4k: 117888 kB\nDirectMap2M: 6168576 kB\nDirectMap1G: 12582912 kB", "capture_warnings": [] }, "summary": { "total": 132, "passed": 107, "failed": 0, "skipped": 25, "timeout": 0, "sub_tests_total": 539, "sub_tests_passed": 457, "sub_tests_failed": 0, "sub_tests_not_tested": 82 }, "results": [ { "name": "GEMM", "category": "kernels", "status": "pass", "duration_sec": 34.82286620140076, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Naive (parallel)", "status": "pass", "max_diff": 1.14e-05, "tolerance": 0.0001, "c_time_us": 1186.5, "pytorch_time_us": 54.1, "speedup": 0.05, "case_id": "GEMM [T,D]x[D,4D] MLP :: M=32, N=512, K=128 :: Naive (parallel)", "configuration": "M=32, N=512, K=128", "evidence_kind": "numerical_and_performance" }, { "name": "AVX-512 (parallel)", "status": "pass", "max_diff": 1.81e-05, "tolerance": 0.0001, "c_time_us": 160.0, "pytorch_time_us": 54.1, "speedup": 0.34, "case_id": "GEMM [T,D]x[D,4D] MLP :: M=32, N=512, K=128 :: AVX-512 (parallel)", "configuration": "M=32, N=512, K=128", "evidence_kind": "numerical_and_performance" }, { "name": "Fine-grained", "status": "pass", "max_diff": 1.72e-05, "tolerance": 0.0001, "c_time_us": 231.4, "pytorch_time_us": 54.1, "speedup": 0.23, "case_id": "GEMM [T,D]x[D,4D] MLP :: M=32, N=512, K=128 :: Fine-grained", "configuration": "M=32, N=512, K=128", "evidence_kind": "numerical_and_performance" }, { "name": "Blocked (serial)", "status": "pass", "max_diff": 1.14e-05, "tolerance": 0.0001, "c_time_us": 1199.7, "pytorch_time_us": 54.1, "speedup": 0.05, "case_id": "GEMM [T,D]x[D,4D] MLP :: M=32, N=512, K=128 :: Blocked (serial)", "configuration": "M=32, N=512, K=128", "evidence_kind": "numerical_and_performance" }, { "name": "Naive (parallel)", "status": "pass", "max_diff": 5.72e-06, "tolerance": 0.0001, "c_time_us": 147.7, "pytorch_time_us": 14.9, "speedup": 0.1, "case_id": "GEMM [T,d]x[T,d] QK^T :: M=64, N=64, K=64 :: Naive (parallel)", "configuration": "M=64, N=64, K=64", "evidence_kind": "numerical_and_performance" }, { "name": "AVX-512 (parallel)", "status": "pass", "max_diff": 1.14e-05, "tolerance": 0.0001, "c_time_us": 24.1, "pytorch_time_us": 14.9, "speedup": 0.62, "case_id": "GEMM [T,d]x[T,d] QK^T :: M=64, N=64, K=64 :: AVX-512 (parallel)", "configuration": "M=64, N=64, K=64", "evidence_kind": "numerical_and_performance" }, { "name": "Fine-grained", "status": "pass", "max_diff": 9.54e-06, "tolerance": 0.0001, "c_time_us": 32.2, "pytorch_time_us": 14.9, "speedup": 0.46, "case_id": "GEMM [T,d]x[T,d] QK^T :: M=64, N=64, K=64 :: Fine-grained", "configuration": "M=64, N=64, K=64", "evidence_kind": "numerical_and_performance" }, { "name": "Blocked (serial)", "status": "pass", "max_diff": 5.72e-06, "tolerance": 0.0001, "c_time_us": 149.3, "pytorch_time_us": 14.9, "speedup": 0.1, "case_id": "GEMM [T,d]x[T,d] QK^T :: M=64, N=64, K=64 :: Blocked (serial)", "configuration": "M=64, N=64, K=64", "evidence_kind": "numerical_and_performance" }, { "name": "Naive (parallel)", "status": "pass", "max_diff": 5.72e-06, "tolerance": 0.0001, "c_time_us": 147.4, "pytorch_time_us": 14.9, "speedup": 0.1, "case_id": "GEMM [T,T]x[T,d] SV :: M=64, N=64, K=64 :: Naive (parallel)", "configuration": "M=64, N=64, K=64", "evidence_kind": "numerical_and_performance" }, { "name": "AVX-512 (parallel)", "status": "pass", "max_diff": 1.14e-05, "tolerance": 0.0001, "c_time_us": 24.4, "pytorch_time_us": 14.9, "speedup": 0.61, "case_id": "GEMM [T,T]x[T,d] SV :: M=64, N=64, K=64 :: AVX-512 (parallel)", "configuration": "M=64, N=64, K=64", "evidence_kind": "numerical_and_performance" }, { "name": "Fine-grained", "status": "pass", "max_diff": 9.54e-06, "tolerance": 0.0001, "c_time_us": 32.7, "pytorch_time_us": 14.9, "speedup": 0.46, "case_id": "GEMM [T,T]x[T,d] SV :: M=64, N=64, K=64 :: Fine-grained", "configuration": "M=64, N=64, K=64", "evidence_kind": "numerical_and_performance" }, { "name": "Blocked (serial)", "status": "pass", "max_diff": 5.72e-06, "tolerance": 0.0001, "c_time_us": 150.3, "pytorch_time_us": 14.9, "speedup": 0.1, "case_id": "GEMM [T,T]x[T,d] SV :: M=64, N=64, K=64 :: Blocked (serial)", "configuration": "M=64, N=64, K=64", "evidence_kind": "numerical_and_performance" }, { "name": "Naive (parallel)", "status": "pass", "max_diff": 4.2e-05, "tolerance": 0.0001, "c_time_us": 25118.0, "pytorch_time_us": 479.2, "speedup": 0.02, "case_id": "GEMM [128,256]x[256,1024] :: M=128, N=1024, K=256 :: Naive (parallel)", "configuration": "M=128, N=1024, K=256", "evidence_kind": "numerical_and_performance" }, { "name": "AVX-512 (parallel)", "status": "pass", "max_diff": 3.43e-05, "tolerance": 0.0001, "c_time_us": 2792.0, "pytorch_time_us": 479.2, "speedup": 0.17, "case_id": "GEMM [128,256]x[256,1024] :: M=128, N=1024, K=256 :: AVX-512 (parallel)", "configuration": "M=128, N=1024, K=256", "evidence_kind": "numerical_and_performance" }, { "name": "Fine-grained", "status": "pass", "max_diff": 3.43e-05, "tolerance": 0.0001, "c_time_us": 4179.4, "pytorch_time_us": 479.2, "speedup": 0.11, "case_id": "GEMM [128,256]x[256,1024] :: M=128, N=1024, K=256 :: Fine-grained", "configuration": "M=128, N=1024, K=256", "evidence_kind": "numerical_and_performance" }, { "name": "Blocked (serial)", "status": "pass", "max_diff": 4.2e-05, "tolerance": 0.0001, "c_time_us": 25125.3, "pytorch_time_us": 479.2, "speedup": 0.02, "case_id": "GEMM [128,256]x[256,1024] :: M=128, N=1024, K=256 :: Blocked (serial)", "configuration": "M=128, N=1024, K=256", "evidence_kind": "numerical_and_performance" } ], "execution_kind": "python", "execution_id": "gemm", "execution_args": [] }, { "name": "GEMM Variants", "category": "kernels", "status": "pass", "duration_sec": 1.529494285583496, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "gemm_variants", "execution_args": [] }, { "name": "GEMM Microkernel", "category": "kernels", "status": "pass", "duration_sec": 17.952858686447144, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "gemm_microkernel", "execution_args": [] }, { "name": "GEMM Fused", "category": "kernels", "status": "pass", "duration_sec": 2.160036563873291, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Tiny (M=8,N=16,K=32)", "status": "pass", "max_diff": 2.86e-06, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Fused GEMM Accuracy (Various Sizes) :: Multiple configurations :: Tiny (M=8,N=16,K=32)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "Small (M=32,N=64,K=128)", "status": "pass", "max_diff": 1.14e-05, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Fused GEMM Accuracy (Various Sizes) :: Multiple configurations :: Small (M=32,N=64,K=128)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "Medium (M=64,N=128,K=256)", "status": "pass", "max_diff": 1.91e-05, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Fused GEMM Accuracy (Various Sizes) :: Multiple configurations :: Medium (M=64,N=128,K=256)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "Large (M=128,N=256,K=512)", "status": "pass", "max_diff": 3.81e-05, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Fused GEMM Accuracy (Various Sizes) :: Multiple configurations :: Large (M=128,N=256,K=512)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "GEMM+Bias+ReLU", "status": "pass", "max_diff": 2.29e-05, "tolerance": 0.0001, "c_time_us": 168.5, "pytorch_time_us": 44.9, "speedup": 0.27, "case_id": "Fused GEMM + Bias + ReLU :: M=64, N=128, K=256 :: GEMM+Bias+ReLU", "configuration": "M=64, N=128, K=256", "evidence_kind": "numerical_and_performance" }, { "name": "GEMM+Bias+GELU", "status": "pass", "max_diff": 0.0207, "tolerance": 0.05, "c_time_us": 254.1, "pytorch_time_us": 88.8, "speedup": 0.35, "case_id": "Fused GEMM + Bias + GELU :: M=64, N=128, K=256 :: GEMM+Bias+GELU", "configuration": "M=64, N=128, K=256", "evidence_kind": "numerical_and_performance" }, { "name": "GEMM+Bias+SiLU", "status": "pass", "max_diff": 1.91e-05, "tolerance": 0.001, "c_time_us": 286.8, "pytorch_time_us": 51.1, "speedup": 0.18, "case_id": "Fused GEMM + Bias + SiLU :: M=64, N=128, K=256 :: GEMM+Bias+SiLU", "configuration": "M=64, N=128, K=256", "evidence_kind": "numerical_and_performance" }, { "name": "GEMM+SwiGLU", "status": "pass", "max_diff": 0.000732, "tolerance": 0.002, "c_time_us": 331.2, "pytorch_time_us": 104.0, "speedup": 0.31, "case_id": "Fused GEMM + SwiGLU (LLaMA MLP) :: M=64, N=128, K=256 :: GEMM+SwiGLU", "configuration": "M=64, N=128, K=256", "evidence_kind": "numerical_and_performance" } ], "execution_kind": "python", "execution_id": "gemm_fused", "execution_args": [] }, { "name": "ReLU", "category": "kernels", "status": "pass", "duration_sec": 2.0141119956970215, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "ReLU (out-of-place)", "status": "pass", "max_diff": 0.0, "tolerance": 1e-07, "c_time_us": 2.5, "pytorch_time_us": 2.9, "speedup": 1.19, "case_id": "ReLU Forward :: N=4096 :: ReLU (out-of-place)", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" }, { "name": "ReLU (inplace)", "status": "pass", "max_diff": 0.0, "tolerance": 1e-07, "c_time_us": 8.1, "pytorch_time_us": 2.9, "speedup": 0.36, "case_id": "ReLU Forward :: N=4096 :: ReLU (inplace)", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" }, { "name": "d_input", "status": "pass", "max_diff": 0.0, "tolerance": 1e-07, "c_time_us": 2.7, "pytorch_time_us": 53.8, "speedup": 19.97, "case_id": "ReLU Backward :: N=4096 :: d_input", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" }, { "name": "Backward", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 2.7, "pytorch_time_us": 50.9, "speedup": 18.88, "case_id": "ReLU Backward :: N=4096 :: Backward", "configuration": "N=4096", "evidence_kind": "performance" } ], "execution_kind": "python", "execution_id": "relu", "execution_args": [] }, { "name": "GELU", "category": "kernels", "status": "pass", "duration_sec": 2.037405490875244, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Fast (SIMD approx)", "status": "pass", "max_diff": 0.0146, "tolerance": 0.02, "c_time_us": 5.2, "pytorch_time_us": 16.7, "speedup": 3.2, "case_id": "GELU Forward :: N=4096 :: Fast (SIMD approx)", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" }, { "name": "Exact (scalar)", "status": "pass", "max_diff": 1.19e-07, "tolerance": 1e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "GELU Forward :: N=4096 :: Exact (scalar)", "configuration": "N=4096", "evidence_kind": "numerical" }, { "name": "GGML (fp16 table)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "GELU Forward :: N=4096 :: GGML (fp16 table)", "configuration": "N=4096", "evidence_kind": "numerical" }, { "name": "SIMD (tanh approx)", "status": "pass", "max_diff": 0.0586, "tolerance": 0.1, "c_time_us": 5.8, "pytorch_time_us": 80.3, "speedup": 13.94, "case_id": "GELU Backward :: N=4096 :: SIMD (tanh approx)", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" }, { "name": "Exact (scalar)", "status": "pass", "max_diff": 4.77e-07, "tolerance": 1e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "GELU Backward :: N=4096 :: Exact (scalar)", "configuration": "N=4096", "evidence_kind": "numerical" }, { "name": "Fast (sigmoid)", "status": "pass", "max_diff": 0.115, "tolerance": 0.1, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "GELU Backward :: N=4096 :: Fast (sigmoid)", "configuration": "N=4096", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "gelu", "execution_args": [] }, { "name": "Sigmoid", "category": "kernels", "status": "pass", "duration_sec": 1.734428882598877, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Sigmoid", "status": "pass", "max_diff": 1.19e-07, "tolerance": 2e-05, "c_time_us": 18.7, "pytorch_time_us": 5.9, "speedup": 0.32, "case_id": "Sigmoid Forward :: N=4096 :: Sigmoid", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" }, { "name": "d_input", "status": "pass", "max_diff": 5.96e-08, "tolerance": 2e-05, "c_time_us": 20.7, "pytorch_time_us": 55.6, "speedup": 2.69, "case_id": "Sigmoid Backward :: N=4096 :: d_input", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" }, { "name": "Backward", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 20.7, "pytorch_time_us": 49.8, "speedup": 2.4, "case_id": "Sigmoid Backward :: N=4096 :: Backward", "configuration": "N=4096", "evidence_kind": "performance" } ], "execution_kind": "python", "execution_id": "sigmoid", "execution_args": [] }, { "name": "Softmax", "category": "kernels", "status": "pass", "duration_sec": 2.468721628189087, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Fast (SIMD approx)", "status": "pass", "max_diff": 0.025, "tolerance": 0.03, "c_time_us": 28.7, "pytorch_time_us": 83.7, "speedup": 2.92, "case_id": "Causal Softmax Forward :: heads=8, tokens=64 :: Fast (SIMD approx)", "configuration": "heads=8, tokens=64", "evidence_kind": "numerical_and_performance" }, { "name": "Exact (scalar)", "status": "pass", "max_diff": 5.96e-08, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Causal Softmax Forward :: heads=8, tokens=64 :: Exact (scalar)", "configuration": "heads=8, tokens=64", "evidence_kind": "numerical" }, { "name": "Fast (SIMD fwd+bwd)", "status": "pass", "max_diff": 0.0348, "tolerance": 0.05, "c_time_us": 44.5, "pytorch_time_us": 224.5, "speedup": 5.05, "case_id": "Causal Softmax Backward :: heads=8, tokens=64 :: Fast (SIMD fwd+bwd)", "configuration": "heads=8, tokens=64", "evidence_kind": "numerical_and_performance" }, { "name": "Exact (scalar fwd+bwd)", "status": "pass", "max_diff": 7.45e-08, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Causal Softmax Backward :: heads=8, tokens=64 :: Exact (scalar fwd+bwd)", "configuration": "heads=8, tokens=64", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "softmax", "execution_args": [] }, { "name": "SwiGLU", "category": "kernels", "status": "pass", "duration_sec": 2.0669214725494385, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Fast (SIMD approx)", "status": "pass", "max_diff": 4.77e-07, "tolerance": 5e-05, "c_time_us": 32.3, "pytorch_time_us": 19.5, "speedup": 0.6, "case_id": "SwiGLU Forward :: T=64, D=128 :: Fast (SIMD approx)", "configuration": "T=64, D=128", "evidence_kind": "numerical_and_performance" }, { "name": "Exact (scalar)", "status": "pass", "max_diff": 4.77e-07, "tolerance": 1e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "SwiGLU Forward :: T=64, D=128 :: Exact (scalar)", "configuration": "T=64, D=128", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "swiglu", "execution_args": [] }, { "name": "LayerNorm", "category": "kernels", "status": "pass", "duration_sec": 1.8564691543579102, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Naive (scalar)", "status": "pass", "max_diff": 1.91e-06, "tolerance": 1e-05, "c_time_us": 11.0, "pytorch_time_us": 6.6, "speedup": 0.6, "case_id": "LayerNorm Forward :: tokens=32, d_model=128 :: Naive (scalar)", "configuration": "tokens=32, d_model=128", "evidence_kind": "numerical_and_performance" }, { "name": "Rolled (SIMD)", "status": "pass", "max_diff": 9.54e-07, "tolerance": 1e-05, "c_time_us": 5.3, "pytorch_time_us": 6.6, "speedup": 1.25, "case_id": "LayerNorm Forward :: tokens=32, d_model=128 :: Rolled (SIMD)", "configuration": "tokens=32, d_model=128", "evidence_kind": "numerical_and_performance" }, { "name": "Unrolled (SIMD)", "status": "pass", "max_diff": 9.54e-07, "tolerance": 1e-05, "c_time_us": 4.3, "pytorch_time_us": 6.6, "speedup": 1.53, "case_id": "LayerNorm Forward :: tokens=32, d_model=128 :: Unrolled (SIMD)", "configuration": "tokens=32, d_model=128", "evidence_kind": "numerical_and_performance" }, { "name": "d_input", "status": "pass", "max_diff": 1.91e-06, "tolerance": 5e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "LayerNorm Backward :: tokens=32, d_model=128 :: d_input", "configuration": "tokens=32, d_model=128", "evidence_kind": "numerical" }, { "name": "d_gamma", "status": "pass", "max_diff": 3.81e-06, "tolerance": 5e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "LayerNorm Backward :: tokens=32, d_model=128 :: d_gamma", "configuration": "tokens=32, d_model=128", "evidence_kind": "numerical" }, { "name": "d_beta", "status": "pass", "max_diff": 1.91e-06, "tolerance": 5e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "LayerNorm Backward :: tokens=32, d_model=128 :: d_beta", "configuration": "tokens=32, d_model=128", "evidence_kind": "numerical" }, { "name": "Forward", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 11.0, "pytorch_time_us": 6.7, "speedup": 0.61, "case_id": "LayerNorm Backward :: tokens=32, d_model=128 :: Forward", "configuration": "tokens=32, d_model=128", "evidence_kind": "performance" }, { "name": "Backward (est)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 11.0, "pytorch_time_us": 80.5, "speedup": 7.29, "case_id": "LayerNorm Backward :: tokens=32, d_model=128 :: Backward (est)", "configuration": "tokens=32, d_model=128", "evidence_kind": "performance" }, { "name": "Forward+Backward", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 21.0, "pytorch_time_us": 87.1, "speedup": 4.15, "case_id": "LayerNorm Backward :: tokens=32, d_model=128 :: Forward+Backward", "configuration": "tokens=32, d_model=128", "evidence_kind": "performance" } ], "execution_kind": "python", "execution_id": "layernorm", "execution_args": [] }, { "name": "RMSNorm", "category": "kernels", "status": "pass", "duration_sec": 2.026601552963257, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "RMSNorm", "status": "pass", "max_diff": 9.54e-07, "tolerance": 1e-06, "c_time_us": 7.0, "pytorch_time_us": 28.1, "speedup": 4.04, "case_id": "RMSNorm Forward :: T=64, D=256 :: RMSNorm", "configuration": "T=64, D=256", "evidence_kind": "numerical_and_performance" }, { "name": "RMSNorm (no rstd cache)", "status": "pass", "max_diff": 9.54e-07, "tolerance": 1e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RMSNorm Forward :: T=64, D=256 :: RMSNorm (no rstd cache)", "configuration": "T=64, D=256", "evidence_kind": "numerical" }, { "name": "d_input", "status": "pass", "max_diff": 1.43e-06, "tolerance": 1e-05, "c_time_us": 14.2, "pytorch_time_us": 220.8, "speedup": 15.59, "case_id": "RMSNorm Backward :: T=64, D=256 :: d_input", "configuration": "T=64, D=256", "evidence_kind": "numerical_and_performance" }, { "name": "d_gamma", "status": "pass", "max_diff": 5.72e-06, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RMSNorm Backward :: T=64, D=256 :: d_gamma", "configuration": "T=64, D=256", "evidence_kind": "numerical" }, { "name": "Backward", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 7.9, "pytorch_time_us": 191.7, "speedup": 24.18, "case_id": "RMSNorm Backward :: T=64, D=256 :: Backward", "configuration": "T=64, D=256", "evidence_kind": "performance" }, { "name": "Forward+Backward", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 14.2, "pytorch_time_us": 220.8, "speedup": 15.59, "case_id": "RMSNorm Backward :: T=64, D=256 :: Forward+Backward", "configuration": "T=64, D=256", "evidence_kind": "performance" } ], "execution_kind": "python", "execution_id": "rmsnorm", "execution_args": [] }, { "name": "RoPE", "category": "kernels", "status": "pass", "duration_sec": 3.04982328414917, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "cos_cache", "status": "pass", "max_diff": 5.96e-08, "tolerance": 5e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Cache Precompute :: max_seq_len=128, head_dim=64 :: cos_cache", "configuration": "max_seq_len=128, head_dim=64", "evidence_kind": "numerical" }, { "name": "sin_cache", "status": "pass", "max_diff": 5.96e-08, "tolerance": 5e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Cache Precompute :: max_seq_len=128, head_dim=64 :: sin_cache", "configuration": "max_seq_len=128, head_dim=64", "evidence_kind": "numerical" }, { "name": "Llama2/Mistral (\u03b8=10000, d=64)", "status": "pass", "max_diff": 5.96e-08, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Cache Multi-Theta Parity :: Various theta values :: Llama2/Mistral (\u03b8=10000, d=64)", "configuration": "Various theta values", "evidence_kind": "numerical" }, { "name": "Llama3 (\u03b8=500000, d=64)", "status": "pass", "max_diff": 5.96e-08, "tolerance": 5e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Cache Multi-Theta Parity :: Various theta values :: Llama3 (\u03b8=500000, d=64)", "configuration": "Various theta values", "evidence_kind": "numerical" }, { "name": "Qwen2 (\u03b8=1000000, d=64)", "status": "pass", "max_diff": 5.96e-08, "tolerance": 5e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Cache Multi-Theta Parity :: Various theta values :: Qwen2 (\u03b8=1000000, d=64)", "configuration": "Various theta values", "evidence_kind": "numerical" }, { "name": "Llama2 (\u03b8=10000)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Forward Multi-Theta :: Various theta values :: Llama2 (\u03b8=10000)", "configuration": "Various theta values", "evidence_kind": "numerical" }, { "name": "Llama3 (\u03b8=500000)", "status": "pass", "max_diff": 4.77e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Forward Multi-Theta :: Various theta values :: Llama3 (\u03b8=500000)", "configuration": "Various theta values", "evidence_kind": "numerical" }, { "name": "Qwen2 (\u03b8=1000000)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Forward Multi-Theta :: Various theta values :: Qwen2 (\u03b8=1000000)", "configuration": "Various theta values", "evidence_kind": "numerical" }, { "name": "Qwen2-1.5B (\u03b8=1000000)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Forward Multi-Theta :: Various theta values :: Qwen2-1.5B (\u03b8=1000000)", "configuration": "Various theta values", "evidence_kind": "numerical" }, { "name": "pos=0", "status": "pass", "max_diff": 0.0, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Decode Positions :: Various positions :: pos=0", "configuration": "Various positions", "evidence_kind": "numerical" }, { "name": "pos=1", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Decode Positions :: Various positions :: pos=1", "configuration": "Various positions", "evidence_kind": "numerical" }, { "name": "pos=10", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Decode Positions :: Various positions :: pos=10", "configuration": "Various positions", "evidence_kind": "numerical" }, { "name": "pos=100", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Decode Positions :: Various positions :: pos=100", "configuration": "Various positions", "evidence_kind": "numerical" }, { "name": "pos=500", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Decode Positions :: Various positions :: pos=500", "configuration": "Various positions", "evidence_kind": "numerical" }, { "name": "pos=1000", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Decode Positions :: Various positions :: pos=1000", "configuration": "Various positions", "evidence_kind": "numerical" }, { "name": "pos=2000", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Decode Positions :: Various positions :: pos=2000", "configuration": "Various positions", "evidence_kind": "numerical" }, { "name": "pairwise-pos=0", "status": "pass", "max_diff": 0.0, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Pairwise Decode Positions :: Various positions / Llama-family :: pairwise-pos=0", "configuration": "Various positions / Llama-family", "evidence_kind": "numerical" }, { "name": "pairwise-pos=1", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Pairwise Decode Positions :: Various positions / Llama-family :: pairwise-pos=1", "configuration": "Various positions / Llama-family", "evidence_kind": "numerical" }, { "name": "pairwise-pos=10", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Pairwise Decode Positions :: Various positions / Llama-family :: pairwise-pos=10", "configuration": "Various positions / Llama-family", "evidence_kind": "numerical" }, { "name": "pairwise-pos=100", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Pairwise Decode Positions :: Various positions / Llama-family :: pairwise-pos=100", "configuration": "Various positions / Llama-family", "evidence_kind": "numerical" }, { "name": "pairwise-pos=500", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Pairwise Decode Positions :: Various positions / Llama-family :: pairwise-pos=500", "configuration": "Various positions / Llama-family", "evidence_kind": "numerical" }, { "name": "pairwise-pos=1000", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Pairwise Decode Positions :: Various positions / Llama-family :: pairwise-pos=1000", "configuration": "Various positions / Llama-family", "evidence_kind": "numerical" }, { "name": "Tiny (H=2,T=8,D=16)", "status": "pass", "max_diff": 1.19e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Accuracy (Various Sizes) :: Multiple configurations :: Tiny (H=2,T=8,D=16)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "Small (H=4,T=16,D=32)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Accuracy (Various Sizes) :: Multiple configurations :: Small (H=4,T=16,D=32)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "Medium (H=8,T=32,D=64)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Accuracy (Various Sizes) :: Multiple configurations :: Medium (H=8,T=32,D=64)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "Large (H=8,T=64,D=128)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Accuracy (Various Sizes) :: Multiple configurations :: Large (H=8,T=64,D=128)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "d_input", "status": "pass", "max_diff": 4.77e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RoPE Backward :: H=8, T=64, D=64 :: d_input", "configuration": "H=8, T=64, D=64", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "rope", "execution_args": [] }, { "name": "Embedding", "category": "kernels", "status": "pass", "duration_sec": 2.1673879623413086, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "d_token_emb", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-06, "c_time_us": 45.9, "pytorch_time_us": 230.2, "speedup": 5.02, "case_id": "Embedding Backward :: V=1000, T=128, D=256 :: d_token_emb", "configuration": "V=1000, T=128, D=256", "evidence_kind": "numerical_and_performance" }, { "name": "d_pos_emb", "status": "pass", "max_diff": 0.0, "tolerance": 1e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Embedding Backward :: V=1000, T=128, D=256 :: d_pos_emb", "configuration": "V=1000, T=128, D=256", "evidence_kind": "numerical" }, { "name": "Backward", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 45.9, "pytorch_time_us": 196.7, "speedup": 4.29, "case_id": "Embedding Backward :: V=1000, T=128, D=256 :: Backward", "configuration": "V=1000, T=128, D=256", "evidence_kind": "performance" } ], "execution_kind": "python", "execution_id": "embedding", "execution_args": [] }, { "name": "Attention", "category": "kernels", "status": "pass", "duration_sec": 2.4145238399505615, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Flash (SIMD)", "status": "pass", "max_diff": 7.75e-07, "tolerance": 0.0001, "c_time_us": 355.5, "pytorch_time_us": 164.7, "speedup": 0.46, "case_id": "Attention Forward (Causal) :: H=8, T=64, D=64 :: Flash (SIMD)", "configuration": "H=8, T=64, D=64", "evidence_kind": "numerical_and_performance" }, { "name": "Score-matrix (exp approx)", "status": "pass", "max_diff": 0.0732, "tolerance": 0.1, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Attention Forward (Causal) :: H=8, T=64, D=64 :: Score-matrix (exp approx)", "configuration": "H=8, T=64, D=64", "evidence_kind": "numerical" }, { "name": "Exact (scalar)", "status": "pass", "max_diff": 4.17e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Attention Forward (Causal) :: H=8, T=64, D=64 :: Exact (scalar)", "configuration": "H=8, T=64, D=64", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "attention", "execution_args": [] }, { "name": "Attention Sliding Window", "category": "kernels", "status": "pass", "duration_sec": 0.12988519668579102, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "prefill_sliding", "status": "pass", "max_diff": 2.98e-08, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "prefill_sliding", "configuration": "", "evidence_kind": "numerical" }, { "name": "prefill_no_window", "status": "pass", "max_diff": 2.235e-08, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "prefill_no_window", "configuration": "", "evidence_kind": "numerical" }, { "name": "decode_sliding", "status": "pass", "max_diff": 2.98e-08, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "decode_sliding", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "attention_sliding", "execution_args": [] }, { "name": "KV Cache Attention", "category": "kernels", "status": "pass", "duration_sec": 1.5385196208953857, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "flash_vs_ref", "status": "pass", "max_diff": 1.31e-06, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention Forward (GQA) :: H=8,H_kv=4,T=64,D=64,aligned=64 :: flash_vs_ref", "configuration": "H=8,H_kv=4,T=64,D=64,aligned=64", "evidence_kind": "numerical" }, { "name": "decode_vs_full", "status": "pass", "max_diff": 9.54e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "KV Cache Decode Attention (GQA) :: H=8,H_kv=4,T=64,D=64,aligned=64 :: decode_vs_full", "configuration": "H=8,H_kv=4,T=64,D=64,aligned=64", "evidence_kind": "numerical" }, { "name": "flash_strided_vs_ref", "status": "pass", "max_diff": 9.54e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention Forward (GQA, Strided KV) :: H=8,H_kv=4,T=64,D=64,aligned=64,stride=128 :: flash_strided_vs_ref", "configuration": "H=8,H_kv=4,T=64,D=64,aligned=64,stride=128", "evidence_kind": "numerical" }, { "name": "flash_vs_ref", "status": "pass", "max_diff": 6.56e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention Forward (GQA) :: H=8,H_kv=4,T=32,D=40,aligned=64 :: flash_vs_ref", "configuration": "H=8,H_kv=4,T=32,D=40,aligned=64", "evidence_kind": "numerical" }, { "name": "decode_vs_full", "status": "pass", "max_diff": 5.96e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "KV Cache Decode Attention (GQA) :: H=8,H_kv=4,T=32,D=40,aligned=64 :: decode_vs_full", "configuration": "H=8,H_kv=4,T=32,D=40,aligned=64", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "kv_cache", "execution_args": [] }, { "name": "KV Cache Decode", "category": "kernels", "status": "pass", "duration_sec": 1.3343186378479004, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "output", "status": "pass", "max_diff": 0.0, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "KV-cache Layer Decode Parity :: output", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "kv_cache_decode", "execution_args": [] }, { "name": "Fused Attention Decode", "category": "kernels", "status": "pass", "duration_sec": 1.3754897117614746, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "fused_attention_decode", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Fused Attention Decode :: decode :: fused_attention_decode", "configuration": "decode", "evidence_kind": "numerical" }, { "name": "fused_attention_decode_mlp", "status": "pass", "max_diff": 0.000427, "tolerance": 0.001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Fused Attention Decode :: decode :: fused_attention_decode_mlp", "configuration": "decode", "evidence_kind": "numerical" }, { "name": "unfused_decode_repeatability", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Fused Attention Decode :: decode :: unfused_decode_repeatability", "configuration": "decode", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "fused_attention_decode", "execution_args": [] }, { "name": "Fused SwiGLU Decode", "category": "kernels", "status": "pass", "duration_sec": 2.8356971740722656, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "mlp_unfused", "status": "pass", "max_diff": 8.34e-07, "tolerance": 0.005, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Fused SwiGLU Decode MLP :: D=896 Hff=4864 :: mlp_unfused", "configuration": "D=896 Hff=4864", "evidence_kind": "numerical" }, { "name": "mlp_fused_v2", "status": "pass", "max_diff": 3.4e-06, "tolerance": 0.005, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Fused SwiGLU Decode MLP :: D=896 Hff=4864 :: mlp_fused_v2", "configuration": "D=896 Hff=4864", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "fused_swiglu_decode", "execution_args": [] }, { "name": "MLP", "category": "kernels", "status": "pass", "duration_sec": 26.620872020721436, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Fast (SIMD GELU)", "status": "pass", "max_diff": 0.00598, "tolerance": 0.01, "c_time_us": 5566.7, "pytorch_time_us": 208.4, "speedup": 0.04, "case_id": "MLP Forward :: T=64, D=128, 4D=512 :: Fast (SIMD GELU)", "configuration": "T=64, D=128, 4D=512", "evidence_kind": "numerical_and_performance" }, { "name": "Exact (scalar GELU)", "status": "pass", "max_diff": 1.64e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "MLP Forward :: T=64, D=128, 4D=512 :: Exact (scalar GELU)", "configuration": "T=64, D=128, 4D=512", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "mlp", "execution_args": [] }, { "name": "Cross Entropy", "category": "kernels", "status": "pass", "duration_sec": 2.3704726696014404, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "d_logits", "status": "pass", "max_diff": 1.16e-10, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Cross-Entropy Loss :: T=64, V=500 :: d_logits", "configuration": "T=64, V=500", "evidence_kind": "numerical" }, { "name": "d_logits", "status": "pass", "max_diff": 5.82e-11, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Cross-Entropy Loss :: T=128, V=1000 :: d_logits", "configuration": "T=128, V=1000", "evidence_kind": "numerical" }, { "name": "Loss (ignore_index=-100)", "status": "pass", "max_diff": 0.0, "tolerance": 1e-05, "c_time_us": 190.1, "pytorch_time_us": 33.6, "speedup": 0.18, "case_id": "Cross-Entropy ignore_index semantics :: T=96, V=503 :: Loss (ignore_index=-100)", "configuration": "T=96, V=503", "evidence_kind": "numerical_and_performance" }, { "name": "d_logits (ignore_index=-100)", "status": "pass", "max_diff": 9.31e-10, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Cross-Entropy ignore_index semantics :: T=96, V=503 :: d_logits (ignore_index=-100)", "configuration": "T=96, V=503", "evidence_kind": "numerical" }, { "name": "Loss (all ignored => NaN)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Cross-Entropy ignore_index semantics :: T=96, V=503 :: Loss (all ignored => NaN)", "configuration": "T=96, V=503", "evidence_kind": "numerical" }, { "name": "d_logits (all ignored => zeros)", "status": "pass", "max_diff": 0.0, "tolerance": 1e-07, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Cross-Entropy ignore_index semantics :: T=96, V=503 :: d_logits (all ignored => zeros)", "configuration": "T=96, V=503", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "cross_entropy", "execution_args": [] }, { "name": "Optimizer (AdamW/SGD)", "category": "training", "status": "pass", "duration_sec": 2.6781017780303955, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "fp32 (step=1)", "status": "pass", "max_diff": 1.19e-07, "tolerance": 1e-06, "c_time_us": 15.8, "pytorch_time_us": 74.5, "speedup": 4.71, "case_id": "AdamW Optimizer :: N=4096 :: fp32 (step=1)", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" }, { "name": "fp32 (10 steps)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "AdamW Optimizer :: N=4096 :: fp32 (10 steps)", "configuration": "N=4096", "evidence_kind": "numerical" }, { "name": "accumulate (+=)", "status": "pass", "max_diff": 0.0, "tolerance": 1e-07, "c_time_us": 7.9, "pytorch_time_us": 2.5, "speedup": 0.32, "case_id": "Gradient Operations :: N=8192 :: accumulate (+=)", "configuration": "N=8192", "evidence_kind": "numerical_and_performance" }, { "name": "scale (*=)", "status": "pass", "max_diff": 0.0, "tolerance": 1e-07, "c_time_us": 5.1, "pytorch_time_us": 2.7, "speedup": 0.54, "case_id": "Gradient Operations :: N=8192 :: scale (*=)", "configuration": "N=8192", "evidence_kind": "numerical_and_performance" } ], "execution_kind": "python", "execution_id": "optimizer", "execution_args": [] }, { "name": "LM Head Litmus", "category": "kernels", "status": "pass", "duration_sec": 22.836017370224, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "lm_head_litmus", "execution_args": [] }, { "name": "DeepSeek Reference Kernels", "category": "kernels", "status": "pass", "duration_sec": 1.697662115097046, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "deepseek_reference", "execution_args": [] }, { "name": "YaRN RoPE Position Providers", "category": "kernels", "status": "pass", "duration_sec": 1.3423199653625488, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "yarn_rope_explicit_positions", "execution_args": [] }, { "name": "ReLU2", "category": "kernels", "status": "pass", "duration_sec": 1.6610119342803955, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "relu2", "execution_args": [] }, { "name": "Nemotron Group-Limited Router", "category": "kernels", "status": "pass", "duration_sec": 1.316246509552002, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "nemotron_router", "execution_args": [] }, { "name": "MoE ReLU2 Expert", "category": "kernels", "status": "pass", "duration_sec": 1.6965277194976807, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "moe_relu2_expert", "execution_args": [] }, { "name": "Mamba2 Reference Kernels", "category": "kernels", "status": "pass", "duration_sec": 1.4108762741088867, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "mamba2_reference", "execution_args": [] }, { "name": "Gemma4 Assistant/MTP Kernels", "category": "kernels", "status": "pass", "duration_sec": 0.17767834663391113, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "gemma4_assistant", "execution_args": [] }, { "name": "Vision", "category": "kernels", "status": "pass", "duration_sec": 2.3997769355773926, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "im2patch (Image 224x224, Patch 16)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 454.0, "pytorch_time_us": 1777.0, "speedup": 3.914096916299559, "case_id": "im2patch (Image 224x224, Patch 16)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "patch2im (Image 224x224, Patch 16)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 87.0, "pytorch_time_us": 800.0, "speedup": 9.195402298850576, "case_id": "patch2im (Image 224x224, Patch 16)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "position_embeddings_add (2304 tokens, dim 1152)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 1956.0, "pytorch_time_us": 1010.0, "speedup": 0.516359918200409, "case_id": "position_embeddings_add (2304 tokens, dim 1152)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "position_embeddings_add_tiled_2d (6x6, dim 8, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 77.0, "pytorch_time_us": null, "speedup": null, "case_id": "position_embeddings_add_tiled_2d (6x6, dim 8, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "position_embeddings_add_tiled_2d Qwen3-VL resize (56x72 from 48x48, dim 1152, merge 2)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "position_embeddings_add_tiled_2d Qwen3-VL resize (56x72 from 48x48, dim 1152, merge 2)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "position_embeddings_add_tiled_2d ragged edges (5x7, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "position_embeddings_add_tiled_2d ragged edges (5x7, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "vision_position_ids_2d_merge (4x4, merge 2)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "vision_position_ids_2d_merge (4x4, merge 2)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "vision_position_ids_2d_merge (6x6, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "vision_position_ids_2d_merge (6x6, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "rowwise_bias_add (2304 tokens, dim 1152)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 1298.0, "pytorch_time_us": 1000.9999999999999, "speedup": 0.771186440677966, "case_id": "rowwise_bias_add (2304 tokens, dim 1152)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "add_stream_reorder_2d (6x6, dim 8, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 101.0, "pytorch_time_us": null, "speedup": null, "case_id": "add_stream_reorder_2d (6x6, dim 8, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "spatial_merge_2x2 (48x48, dim 64)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 273.0, "pytorch_time_us": 221.0, "speedup": 0.8095238095238094, "case_id": "spatial_merge_2x2 (48x48, dim 64)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "spatial_merge_contiguous_tiled (6x6, dim 5, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 37.0, "pytorch_time_us": null, "speedup": null, "case_id": "spatial_merge_contiguous_tiled (6x6, dim 5, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "feature_slice_copy (6 rows, src 4, dst 11, off 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 47.0, "pytorch_time_us": null, "speedup": null, "case_id": "feature_slice_copy (6 rows, src 4, dst 11, off 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "feature_concat (5 rows, main 3, branch 2 x 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 91.0, "pytorch_time_us": null, "speedup": null, "case_id": "feature_concat (5 rows, main 3, branch 2 x 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "feature_concat in-place expand (5 rows, main 3, branch 2 x 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "feature_concat in-place expand (5 rows, main 3, branch 2 x 3)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 8) Q", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 8) Q", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 8) K", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 8) K", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 72) Q", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 72) Q", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 72) K", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 72) K", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "im2patch (Image 32x32, Patch 8)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 46.0, "pytorch_time_us": 324.0, "speedup": 7.043478260869565, "case_id": "im2patch (Image 32x32, Patch 8)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "patch2im (Image 32x32, Patch 8)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 40.0, "pytorch_time_us": 87.0, "speedup": 2.175, "case_id": "patch2im (Image 32x32, Patch 8)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" } ], "execution_kind": "python", "execution_id": "vision", "execution_args": [] }, { "name": "Audio Log-Mel Frontend", "category": "kernels", "status": "pass", "duration_sec": 4.939889907836914, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "whisper_log_mel_tones", "status": "pass", "max_diff": 0.00106799603, "tolerance": 0.0011, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "whisper_log_mel_tones", "configuration": "", "evidence_kind": "numerical" }, { "name": "whisper_log_mel_noise", "status": "pass", "max_diff": 6.7949295e-06, "tolerance": 0.0011, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "whisper_log_mel_noise", "configuration": "", "evidence_kind": "numerical" }, { "name": "whisper_log_mel_impulse", "status": "pass", "max_diff": 3.57627869e-07, "tolerance": 0.0011, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "whisper_log_mel_impulse", "configuration": "", "evidence_kind": "numerical" }, { "name": "whisper_log_mel_stage_composition", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "whisper_log_mel_stage_composition", "configuration": "", "evidence_kind": "numerical" }, { "name": "whisper_log_mel_shape_contract", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "whisper_log_mel_shape_contract", "configuration": "", "evidence_kind": "numerical" }, { "name": "whisper_slaney_80_filter_identity", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "whisper_slaney_80_filter_identity", "configuration": "", "evidence_kind": "numerical" }, { "name": "whisper_log_mel_30s_silence(shape=80x3000)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "whisper_log_mel_30s_silence(shape=80x3000)", "configuration": "", "evidence_kind": "numerical" }, { "name": "whisper_global_feature_window", "status": "pass", "max_diff": 0.00016283989, "tolerance": 0.0011, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "whisper_global_feature_window", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "audio_frontend", "execution_args": [] }, { "name": "Audio Transformer Primitives", "category": "kernels", "status": "pass", "duration_sec": 5.238799333572388, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "audio_wav_pcm16_chunked_decode", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_wav_pcm16_chunked_decode", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_pcm_s16_stereo_to_mono", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_pcm_s16_stereo_to_mono", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_pad_or_truncate", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_pad_or_truncate", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_hann_window", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_hann_window", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_preemphasis", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_preemphasis", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_relative_sinusoidal_position", "status": "pass", "max_diff": 5.96046448e-08, "tolerance": 6e-08, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_relative_sinusoidal_position", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_batch_norm_inference", "status": "pass", "max_diff": 1.1920929e-07, "tolerance": 2.5e-07, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_batch_norm_inference", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_lstm_step", "status": "pass", "max_diff": 1.49011612e-08, "tolerance": 1e-07, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_lstm_step", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_per_feature_normalization", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_per_feature_normalization", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_resample_linear_48k_to_16k", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_resample_linear_48k_to_16k", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_resample_linear_44k1_to_16k", "status": "pass", "max_diff": 1.49011612e-08, "tolerance": 1e-07, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_resample_linear_44k1_to_16k", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_resample_windowed_sinc", "status": "pass", "max_diff": 0.0, "tolerance": 3e-08, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_resample_windowed_sinc", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_stft_precomputed_vs_direct", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_stft_precomputed_vs_direct", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_stft_fft400_vs_direct", "status": "pass", "max_diff": 0.000199794769, "tolerance": 0.0004, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_stft_fft400_vs_direct", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_centered_window_stft", "status": "pass", "max_diff": 5.96046448e-08, "tolerance": 2e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_centered_window_stft", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_fastconformer_subsampling", "status": "pass", "max_diff": 0.0, "tolerance": 3e-07, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_fastconformer_subsampling", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_fastconformer_block", "status": "pass", "max_diff": 2.38418579e-07, "tolerance": 2e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_fastconformer_block", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_split_glu", "status": "pass", "max_diff": 0.0, "tolerance": 1.2e-07, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_split_glu", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_conformer_relative_attention", "status": "pass", "max_diff": 1.49011612e-08, "tolerance": 1.5e-07, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_conformer_relative_attention", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_conv1d_whisper_stem1", "status": "pass", "max_diff": 4.17232513e-07, "tolerance": 2e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_conv1d_whisper_stem1", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_conv1d_whisper_stem2", "status": "pass", "max_diff": 3.57627869e-07, "tolerance": 2e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_conv1d_whisper_stem2", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_erf_gelu_fp64_scalar", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_erf_gelu_fp64_scalar", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_erf_gelu_fp64_parallel", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_erf_gelu_fp64_parallel", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_channel_to_token_transpose", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_channel_to_token_transpose", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_encoder_self_attention_equal", "status": "pass", "max_diff": 2.23517418e-08, "tolerance": 2e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_encoder_self_attention_equal", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_cross_attention_unequal_small", "status": "pass", "max_diff": 1.11758709e-08, "tolerance": 2e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_cross_attention_unequal_small", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_cross_attention_whisper_decode", "status": "pass", "max_diff": 5.82076609e-09, "tolerance": 2e-06, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_cross_attention_whisper_decode", "configuration": "", "evidence_kind": "numerical" }, { "name": "audio_encoder_tiled_f16kv_attention", "status": "pass", "max_diff": 2.14204192e-08, "tolerance": 2e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "audio_encoder_tiled_f16kv_attention", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "audio_transformer_primitives", "execution_args": [] }, { "name": "Nightly Runner Hardware Capture", "category": "inference", "status": "pass", "duration_sec": 0.07314181327819824, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "nightly_runner_hardware", "execution_args": [] }, { "name": "Idle-host Nightly Coordinator", "category": "inference", "status": "pass", "duration_sec": 0.5464472770690918, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "idle_nightly_coordinator", "execution_args": [] }, { "name": "ReLU BF16", "category": "bf16", "status": "pass", "duration_sec": 1.7844018936157227, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "ReLU", "status": "pass", "max_diff": 0.0, "tolerance": 0.01, "c_time_us": 2.2, "pytorch_time_us": 3.6, "speedup": 1.62, "case_id": "ReLU Forward (BF16) :: N=4096 :: ReLU", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" }, { "name": "d_input", "status": "pass", "max_diff": 0.0, "tolerance": 0.01, "c_time_us": 2.4, "pytorch_time_us": 54.1, "speedup": 22.17, "case_id": "ReLU Backward (BF16) :: N=4096 :: d_input", "configuration": "N=4096", "evidence_kind": "numerical_and_performance" } ], "execution_kind": "python", "execution_id": "relu_bf16", "execution_args": [] }, { "name": "GELU BF16", "category": "bf16", "status": "skip", "duration_sec": 1.2990007400512695, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "gelu_bf16", "execution_args": [] }, { "name": "Sigmoid BF16", "category": "bf16", "status": "skip", "duration_sec": 1.3016841411590576, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "sigmoid_bf16", "execution_args": [] }, { "name": "SwiGLU BF16", "category": "bf16", "status": "skip", "duration_sec": 1.3647148609161377, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "swiglu_bf16", "execution_args": [] }, { "name": "LayerNorm BF16", "category": "bf16", "status": "skip", "duration_sec": 1.3854529857635498, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "layernorm_bf16", "execution_args": [] }, { "name": "RMSNorm BF16", "category": "bf16", "status": "skip", "duration_sec": 1.3594036102294922, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "rmsnorm_bf16", "execution_args": [] }, { "name": "RoPE BF16", "category": "bf16", "status": "skip", "duration_sec": 1.3525950908660889, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "rope_bf16", "execution_args": [] }, { "name": "YaRN RoPE Position Providers BF16", "category": "bf16", "status": "pass", "duration_sec": 1.3197674751281738, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "yarn_rope_explicit_positions_bf16", "execution_args": [] }, { "name": "Embedding BF16", "category": "bf16", "status": "skip", "duration_sec": 1.3649609088897705, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "embedding_bf16", "execution_args": [] }, { "name": "Attention BF16", "category": "bf16", "status": "skip", "duration_sec": 1.415217638015747, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "attention_bf16", "execution_args": [] }, { "name": "MLP BF16", "category": "bf16", "status": "skip", "duration_sec": 1.3110716342926025, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "mlp_bf16", "execution_args": [] }, { "name": "Cross Entropy BF16", "category": "bf16", "status": "skip", "duration_sec": 1.3084580898284912, "stdout": "", "stderr": "", "error_msg": "BF16 kernels require AVX-512 BF16; skipping this test on the current CPU.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "cross_entropy_bf16", "execution_args": [] }, { "name": "Vision Position Storage BF16", "category": "bf16", "status": "pass", "duration_sec": 1.3833510875701904, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "vision_position_storage_bf16", "execution_args": [] }, { "name": "v8 BF16 safetensors lowering", "category": "bf16", "status": "pass", "duration_sec": 7.417383432388306, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 72) Q", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 72) Q", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 72) K", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 72) K", "configuration": "", "evidence_kind": "numerical_measurement" } ], "execution_kind": "python", "execution_id": "v8_bf16_safetensors_lowering", "execution_args": [] }, { "name": "v8 BF16 file-backed BUMP", "category": "bf16", "status": "pass", "duration_sec": 0.12366342544555664, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "v8_bf16_file_backed_bump", "execution_args": [] }, { "name": "Instella-MoE Circuit/Codegen Contracts", "category": "inference", "status": "pass", "duration_sec": 5.174670457839966, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "v8_instella_moe_circuit_contracts", "execution_args": [] }, { "name": "PyTorch/oneDNN Oracle Contract BF16", "category": "bf16", "status": "pass", "duration_sec": 0.1354844570159912, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "pytorch_onednn_oracle_contract_bf16", "execution_args": [] }, { "name": "Q4_K Kernels", "category": "quant", "status": "pass", "duration_sec": 1.386561393737793, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Q4_K Dequantization", "status": "pass", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Q4_K Dequantization", "configuration": "", "evidence_kind": "numerical" }, { "name": "Q4_K GEMV", "status": "pass", "max_diff": 0.00061, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Q4_K GEMV", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "q4k_kernels", "execution_args": [] }, { "name": "Q6_K Kernels", "category": "quant", "status": "pass", "duration_sec": 1.3685529232025146, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "dequant_q6_k_row", "status": "pass", "max_diff": 0.0, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Q6_K Kernels :: block=256 :: dequant_q6_k_row", "configuration": "block=256", "evidence_kind": "numerical" }, { "name": "gemv_q6_k (M=32,K=256)", "status": "pass", "max_diff": 7.63e-05, "tolerance": 0.0005, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Q6_K Kernels :: block=256 :: gemv_q6_k (M=32,K=256)", "configuration": "block=256", "evidence_kind": "numerical" }, { "name": "gemm_nt_q6_k (M=1,N=32,K=256)", "status": "pass", "max_diff": 3.05e-05, "tolerance": 0.0005, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Q6_K Kernels :: block=256 :: gemm_nt_q6_k (M=1,N=32,K=256)", "configuration": "block=256", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "q6k_kernels", "execution_args": [] }, { "name": "Q4_K x Q8_K MatVec", "category": "quant", "status": "pass", "duration_sec": 1.4755048751831055, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "End-to-end (M=16,K=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": 3.8, "pytorch_time_us": 3.3, "speedup": 0.86, "case_id": "Q4_K x Q8_K Matvec (Performance) :: M=16, K=512, y=W@x :: End-to-end (M=16,K=512)", "configuration": "M=16, K=512, y=W@x", "evidence_kind": "numerical_and_performance" }, { "name": "End-to-end (M=64,K=2048)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": 9.6, "pytorch_time_us": 16.2, "speedup": 1.69, "case_id": "Q4_K x Q8_K Matvec (Performance) :: M=64, K=2048, y=W@x :: End-to-end (M=64,K=2048)", "configuration": "M=64, K=2048, y=W@x", "evidence_kind": "numerical_and_performance" }, { "name": "End-to-end (M=16,K=1024)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": 4.4, "pytorch_time_us": 3.9, "speedup": 0.88, "case_id": "Q4_K x Q8_K Matvec (Performance) :: M=16, K=896 (padded 1024), y=W@x :: End-to-end (M=16,K=1024)", "configuration": "M=16, K=896 (padded 1024), y=W@x", "evidence_kind": "numerical_and_performance" } ], "execution_kind": "python", "execution_id": "q4_k_q8_k_matvec", "execution_args": [] }, { "name": "Softmax Backward", "category": "training", "status": "pass", "duration_sec": 1.4478588104248047, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Small (H=2,T=8)", "status": "pass", "max_diff": 2.98e-08, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Causal Softmax Backward Accuracy :: Multiple sizes :: Small (H=2,T=8)", "configuration": "Multiple sizes", "evidence_kind": "numerical" }, { "name": "Medium (H=4,T=16)", "status": "pass", "max_diff": 1.19e-07, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Causal Softmax Backward Accuracy :: Multiple sizes :: Medium (H=4,T=16)", "configuration": "Multiple sizes", "evidence_kind": "numerical" }, { "name": "Large (H=8,T=32)", "status": "pass", "max_diff": 5.96e-08, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Causal Softmax Backward Accuracy :: Multiple sizes :: Large (H=8,T=32)", "configuration": "Multiple sizes", "evidence_kind": "numerical" }, { "name": "d_scores", "status": "pass", "max_diff": 5.96e-08, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Causal Softmax Backward :: H=8, T=64 :: d_scores", "configuration": "H=8, T=64", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "softmax_backward", "execution_args": [] }, { "name": "Attention Backward", "category": "training", "status": "pass", "duration_sec": 16.068207025527954, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Tiny (H=2,T=8,D=16)", "status": "pass", "max_diff": 4.77e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Attention Backward Accuracy :: Multiple configurations :: Tiny (H=2,T=8,D=16)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "Small (H=4,T=16,D=32)", "status": "pass", "max_diff": 9.54e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Attention Backward Accuracy :: Multiple configurations :: Small (H=4,T=16,D=32)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "Medium (H=8,T=32,D=64)", "status": "pass", "max_diff": 1.19e-06, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Attention Backward Accuracy :: Multiple configurations :: Medium (H=8,T=32,D=64)", "configuration": "Multiple configurations", "evidence_kind": "numerical" }, { "name": "d_q", "status": "pass", "max_diff": 4.77e-07, "tolerance": 0.0001, "c_time_us": 141.6, "pytorch_time_us": 21903.5, "speedup": 154.66, "case_id": "Attention Backward (Non-GQA) :: H=4, T=32, D=32 :: d_q", "configuration": "H=4, T=32, D=32", "evidence_kind": "numerical_and_performance" }, { "name": "d_k", "status": "pass", "max_diff": 5.96e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Attention Backward (Non-GQA) :: H=4, T=32, D=32 :: d_k", "configuration": "H=4, T=32, D=32", "evidence_kind": "numerical" }, { "name": "d_v", "status": "pass", "max_diff": 9.54e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Attention Backward (Non-GQA) :: H=4, T=32, D=32 :: d_v", "configuration": "H=4, T=32, D=32", "evidence_kind": "numerical" }, { "name": "d_q (GQA)", "status": "pass", "max_diff": 7.15e-07, "tolerance": 0.0001, "c_time_us": 236.0, "pytorch_time_us": 45123.5, "speedup": 191.18, "case_id": "Attention Backward (GQA) :: H=8, H_kv=2, T=32, D=32 :: d_q (GQA)", "configuration": "H=8, H_kv=2, T=32, D=32", "evidence_kind": "numerical_and_performance" }, { "name": "d_k (GQA)", "status": "pass", "max_diff": 1.19e-06, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Attention Backward (GQA) :: H=8, H_kv=2, T=32, D=32 :: d_k (GQA)", "configuration": "H=8, H_kv=2, T=32, D=32", "evidence_kind": "numerical" }, { "name": "d_v (GQA)", "status": "pass", "max_diff": 2.86e-06, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Attention Backward (GQA) :: H=8, H_kv=2, T=32, D=32 :: d_v (GQA)", "configuration": "H=8, H_kv=2, T=32, D=32", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "attention_backward", "execution_args": [] }, { "name": "DeltaNet Backward", "category": "training", "status": "pass", "duration_sec": 1.7843363285064697, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "deltanet_backward", "execution_args": [] }, { "name": "PyTorch Parity", "category": "parity", "status": "pass", "duration_sec": 1.8379004001617432, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Forward", "status": "pass", "max_diff": 1.43e-06, "tolerance": 1e-05, "c_time_us": 15.3, "pytorch_time_us": 199.5, "speedup": 13.01, "case_id": "RMSNorm Forward/Backward Parity :: T=64, D=128 :: Forward", "configuration": "T=64, D=128", "evidence_kind": "numerical_and_performance" }, { "name": "d_input", "status": "pass", "max_diff": 9.54e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RMSNorm Forward/Backward Parity :: T=64, D=128 :: d_input", "configuration": "T=64, D=128", "evidence_kind": "numerical" }, { "name": "d_gamma", "status": "pass", "max_diff": 3.81e-06, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RMSNorm Forward/Backward Parity :: T=64, D=128 :: d_gamma", "configuration": "T=64, D=128", "evidence_kind": "numerical" }, { "name": "Forward", "status": "pass", "max_diff": 4.77e-07, "tolerance": 5e-05, "c_time_us": 39.1, "pytorch_time_us": 24.8, "speedup": 0.63, "case_id": "SwiGLU Forward/Backward Parity :: T=64, D=128 :: Forward", "configuration": "T=64, D=128", "evidence_kind": "numerical_and_performance" }, { "name": "d_input", "status": "pass", "max_diff": 3.79e-05, "tolerance": 5e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "SwiGLU Forward/Backward Parity :: T=64, D=128 :: d_input", "configuration": "T=64, D=128", "evidence_kind": "numerical" }, { "name": "Loss", "status": "pass", "max_diff": 0.0, "tolerance": 1e-05, "c_time_us": 170.3, "pytorch_time_us": 27.9, "speedup": 0.16, "case_id": "Cross-Entropy Loss/Gradient Parity :: T=64, V=500 :: Loss", "configuration": "T=64, V=500", "evidence_kind": "numerical_and_performance" }, { "name": "d_logits", "status": "pass", "max_diff": 9.31e-10, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Cross-Entropy Loss/Gradient Parity :: T=64, V=500 :: d_logits", "configuration": "T=64, V=500", "evidence_kind": "numerical" } ], "execution_kind": "python", "execution_id": "pytorch_parity", "execution_args": [] }, { "name": "Qwen3-VL Private Corpus Runner Contract", "category": "parity", "status": "pass", "duration_sec": 0.14058184623718262, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "qwen3vl_private_corpus_contract", "execution_args": [] }, { "name": "v8 Generated Layout Bridge Execution", "category": "parity", "status": "pass", "duration_sec": 0.6588244438171387, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "v8_generated_layout_bridges", "execution_args": [] }, { "name": "RoPE Pairwise Layout (Llama)", "category": "parity", "status": "pass", "duration_sec": 0.17066645622253418, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "python", "execution_id": "rope_pairwise_layout", "execution_args": [] }, { "name": "v8 Native Session/CLI FFI", "category": "inference", "status": "pass", "duration_sec": 9.833801507949829, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-native-session-v8", "execution_args": [] }, { "name": "Experimental Responses Server Schema", "category": "inference", "status": "pass", "duration_sec": 4.8081512451171875, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-server-schema", "execution_args": [] }, { "name": "v8 Audio Circuit/Codegen Contracts", "category": "inference", "status": "skip", "duration_sec": 36.628509759902954, "stdout": "", "stderr": "", "error_msg": "SKIP: set CK_WHISPER_CHECKPOINT, CK_WHISPER_ENCODER_RUN_DIR, CK_WHISPER_DECODER_RUN_DIR, and CK_WHISPER_WAV", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-audio-v8-contracts", "execution_args": [] }, { "name": "Tokenizer Byte/UTF-8 Parity", "category": "inference", "status": "pass", "duration_sec": 7.294989824295044, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-tokenizer-special", "execution_args": [] }, { "name": "Litmus Test", "category": "parity", "status": "pass", "duration_sec": 23.76984477043152, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "litmus", "execution_args": [] }, { "name": "llama.cpp Parity (Nightly)", "category": "parity", "status": "pass", "duration_sec": 172.78218603134155, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "prefill_sliding", "status": "pass", "max_diff": 2.98e-08, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "prefill_sliding", "configuration": "", "evidence_kind": "numerical" }, { "name": "prefill_no_window", "status": "pass", "max_diff": 2.235e-08, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "prefill_no_window", "configuration": "", "evidence_kind": "numerical" }, { "name": "decode_sliding", "status": "pass", "max_diff": 2.98e-08, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "decode_sliding", "configuration": "", "evidence_kind": "numerical" }, { "name": "Forward", "status": "pass", "max_diff": 1.43e-06, "tolerance": 1e-05, "c_time_us": 15.6, "pytorch_time_us": 204.2, "speedup": 13.07, "case_id": "RMSNorm Forward/Backward Parity :: T=64, D=128 :: Forward", "configuration": "T=64, D=128", "evidence_kind": "numerical_and_performance" }, { "name": "d_input", "status": "pass", "max_diff": 9.54e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RMSNorm Forward/Backward Parity :: T=64, D=128 :: d_input", "configuration": "T=64, D=128", "evidence_kind": "numerical" }, { "name": "d_gamma", "status": "pass", "max_diff": 3.81e-06, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "RMSNorm Forward/Backward Parity :: T=64, D=128 :: d_gamma", "configuration": "T=64, D=128", "evidence_kind": "numerical" }, { "name": "Forward", "status": "pass", "max_diff": 4.77e-07, "tolerance": 5e-05, "c_time_us": 38.5, "pytorch_time_us": 23.8, "speedup": 0.62, "case_id": "SwiGLU Forward/Backward Parity :: T=64, D=128 :: Forward", "configuration": "T=64, D=128", "evidence_kind": "numerical_and_performance" }, { "name": "d_input", "status": "pass", "max_diff": 3.79e-05, "tolerance": 5e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "SwiGLU Forward/Backward Parity :: T=64, D=128 :: d_input", "configuration": "T=64, D=128", "evidence_kind": "numerical" }, { "name": "Loss", "status": "pass", "max_diff": 0.0, "tolerance": 1e-05, "c_time_us": 166.7, "pytorch_time_us": 26.5, "speedup": 0.16, "case_id": "Cross-Entropy Loss/Gradient Parity :: T=64, V=500 :: Loss", "configuration": "T=64, V=500", "evidence_kind": "numerical_and_performance" }, { "name": "d_logits", "status": "pass", "max_diff": 9.31e-10, "tolerance": 1e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Cross-Entropy Loss/Gradient Parity :: T=64, V=500 :: d_logits", "configuration": "T=64, V=500", "evidence_kind": "numerical" } ], "execution_kind": "make", "execution_id": "llamacpp-parity-nightly", "execution_args": [] }, { "name": "Flash Attention (50K+)", "category": "kernels", "status": "pass", "duration_sec": 16.790029764175415, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "fast_exp_vec=avx", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention - Config :: tile_k_max=32, tile_heuristic=on :: fast_exp_vec=avx", "configuration": "tile_k_max=32, tile_heuristic=on", "evidence_kind": "numerical" }, { "name": "decode_small (Tq=1,Tk=16,H=2,D=32)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_small (Tq=1,Tk=16,H=2,D=32)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "fastexp_decode_small (Tq=1,Tk=16,H=2,D=32)", "status": "pass", "max_diff": 1.93e-05, "tolerance": 0.01, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: fastexp_decode_small (Tq=1,Tk=16,H=2,D=32)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "decode_tail (Tq=1,Tk=63,H=4,D=40)", "status": "pass", "max_diff": 2.68e-07, "tolerance": 0.0002, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_tail (Tq=1,Tk=63,H=4,D=40)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "fastexp_decode_tail (Tq=1,Tk=63,H=4,D=40)", "status": "pass", "max_diff": 7.79e-06, "tolerance": 0.01, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: fastexp_decode_tail (Tq=1,Tk=63,H=4,D=40)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "prefill (Tq=8,Tk=8,H=4,D=64)", "status": "pass", "max_diff": 3.58e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: prefill (Tq=8,Tk=8,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "fastexp_prefill (Tq=8,Tk=8,H=4,D=64)", "status": "pass", "max_diff": 1.9e-05, "tolerance": 0.01, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: fastexp_prefill (Tq=8,Tk=8,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "offset (Tq=3,Tk=9,H=3,D=24)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: offset (Tq=3,Tk=9,H=3,D=24)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "fastexp_offset (Tq=3,Tk=9,H=3,D=24)", "status": "pass", "max_diff": 1.03e-05, "tolerance": 0.01, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: fastexp_offset (Tq=3,Tk=9,H=3,D=24)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "prefill_hd128 (Tq=4,Tk=4,H=2,D=128)", "status": "pass", "max_diff": 4.77e-07, "tolerance": 0.0003, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: prefill_hd128 (Tq=4,Tk=4,H=2,D=128)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "fastexp_prefill_hd128 (Tq=4,Tk=4,H=2,D=128)", "status": "pass", "max_diff": 1.65e-05, "tolerance": 0.01, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: fastexp_prefill_hd128 (Tq=4,Tk=4,H=2,D=128)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "decode_hd192 (Tq=1,Tk=6,H=2,D=192)", "status": "pass", "max_diff": 4.77e-07, "tolerance": 0.0004, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_hd192 (Tq=1,Tk=6,H=2,D=192)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "fastexp_decode_hd192 (Tq=1,Tk=6,H=2,D=192)", "status": "pass", "max_diff": 1.07e-05, "tolerance": 0.01, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: fastexp_decode_hd192 (Tq=1,Tk=6,H=2,D=192)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "decode_ln_gelu (Tq=1,Tk=32,H=4,D=64,dist=ln_gelu)", "status": "pass", "max_diff": 2.09e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_ln_gelu (Tq=1,Tk=32,H=4,D=64,dist=ln_gelu)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "fastexp_decode_ln_gelu (Tq=1,Tk=32,H=4,D=64,dist=ln_gelu)", "status": "pass", "max_diff": 5.48e-06, "tolerance": 0.01, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: fastexp_decode_ln_gelu (Tq=1,Tk=32,H=4,D=64,dist=ln_gelu)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "prefill_gqa_8h_2kv (Tq=4,Tk=4,H=8,D=64)", "status": "pass", "max_diff": 4.02e-07, "tolerance": 0.0002, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: prefill_gqa_8h_2kv (Tq=4,Tk=4,H=8,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "fastexp_prefill_gqa_8h_2kv (Tq=4,Tk=4,H=8,D=64)", "status": "pass", "max_diff": 3.49e-05, "tolerance": 0.01, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: fastexp_prefill_gqa_8h_2kv (Tq=4,Tk=4,H=8,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "decode_wrapper_omp (Tq=1,Tk=64,H=4,D=64)", "status": "pass", "max_diff": 2.68e-07, "tolerance": 0.0001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_wrapper_omp (Tq=1,Tk=64,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "decode_gqa_8h_2kv (Tq=1,Tk=64,H=8,D=64,kv=2)", "status": "pass", "max_diff": 2.38e-07, "tolerance": 0.0002, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_gqa_8h_2kv (Tq=1,Tk=64,H=8,D=64,kv=2)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "decode_mqa_16h_1kv (Tq=1,Tk=64,H=16,D=64,kv=1)", "status": "pass", "max_diff": 2.98e-07, "tolerance": 0.0002, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_mqa_16h_1kv (Tq=1,Tk=64,H=16,D=64,kv=1)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "llama.cpp (decode_small)", "status": "pass", "max_diff": 3.58e-07, "tolerance": 0.001, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: llama.cpp (decode_small)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "numerical" }, { "name": "decode_4h_512 (Tq=1,Tk=512,H=4,D=64)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 32.6, "pytorch_time_us": 24.7, "speedup": 0.76, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_4h_512 (Tq=1,Tk=512,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "performance" }, { "name": "decode_8h_512 (Tq=1,Tk=512,H=8,D=64)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 39.1, "pytorch_time_us": 46.5, "speedup": 1.19, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_8h_512 (Tq=1,Tk=512,H=8,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "performance" }, { "name": "decode_32h_512 (Tq=1,Tk=512,H=32,D=64)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 102.1, "pytorch_time_us": 174.8, "speedup": 1.71, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_32h_512 (Tq=1,Tk=512,H=32,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "performance" }, { "name": "decode_4h_1k (Tq=1,Tk=1024,H=4,D=64)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 32.1, "pytorch_time_us": 44.6, "speedup": 1.39, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_4h_1k (Tq=1,Tk=1024,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "performance" }, { "name": "decode_4h_8k (Tq=1,Tk=8192,H=4,D=64)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 180.8, "pytorch_time_us": 320.9, "speedup": 1.77, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_4h_8k (Tq=1,Tk=8192,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "performance" }, { "name": "decode_4h_24k (Tq=1,Tk=24576,H=4,D=64)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 1324.9, "pytorch_time_us": 1068.2, "speedup": 0.81, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_4h_24k (Tq=1,Tk=24576,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "performance" }, { "name": "decode_4h_50k (Tq=1,Tk=51200,H=4,D=64)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 2495.4, "pytorch_time_us": 2378.7, "speedup": 0.95, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_4h_50k (Tq=1,Tk=51200,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "performance" }, { "name": "decode_4h_hd128 (Tq=1,Tk=256,H=4,D=128)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 22.5, "pytorch_time_us": 18.4, "speedup": 0.82, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_4h_hd128 (Tq=1,Tk=256,H=4,D=128)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "performance" }, { "name": "decode_32h_51200 (Tq=1,Tk=51200,H=32,D=64)", "status": "not_tested", "max_diff": null, "tolerance": null, "c_time_us": 21081.3, "pytorch_time_us": 41665.4, "speedup": 1.98, "case_id": "Flash Attention (Causal, Online Softmax) - Accuracy :: tile_k_max=32, tile_heuristic=on, fast_exp=off :: decode_32h_51200 (Tq=1,Tk=51200,H=32,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=off", "evidence_kind": "performance" }, { "name": "decode_4h_512 (Tq=1,Tk=512,H=4,D=64)", "status": "pass", "max_diff": 3.29e-06, "tolerance": 0.001, "c_time_us": 32.6, "pytorch_time_us": 111.0, "speedup": 3.4, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_4h_512 (Tq=1,Tk=512,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_8h_512 (Tq=1,Tk=512,H=8,D=64)", "status": "pass", "max_diff": 3.6e-06, "tolerance": 0.001, "c_time_us": 39.1, "pytorch_time_us": 130.4, "speedup": 3.33, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_8h_512 (Tq=1,Tk=512,H=8,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_32h_512 (Tq=1,Tk=512,H=32,D=64)", "status": "pass", "max_diff": 4.7e-06, "tolerance": 0.001, "c_time_us": 102.1, "pytorch_time_us": 306.6, "speedup": 3.0, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_32h_512 (Tq=1,Tk=512,H=32,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_4h_1k (Tq=1,Tk=1024,H=4,D=64)", "status": "pass", "max_diff": 2.09e-06, "tolerance": 0.001, "c_time_us": 32.1, "pytorch_time_us": 115.2, "speedup": 3.58, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_4h_1k (Tq=1,Tk=1024,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_4h_8k (Tq=1,Tk=8192,H=4,D=64)", "status": "pass", "max_diff": 7.89e-07, "tolerance": 0.001, "c_time_us": 180.8, "pytorch_time_us": 474.4, "speedup": 2.62, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_4h_8k (Tq=1,Tk=8192,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_4h_24k (Tq=1,Tk=24576,H=4,D=64)", "status": "pass", "max_diff": 5.9e-07, "tolerance": 0.001, "c_time_us": 1324.9, "pytorch_time_us": 2871.7, "speedup": 2.17, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_4h_24k (Tq=1,Tk=24576,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_4h_50k (Tq=1,Tk=51200,H=4,D=64)", "status": "pass", "max_diff": 3.3e-07, "tolerance": 0.001, "c_time_us": 2495.4, "pytorch_time_us": 6331.0, "speedup": 2.54, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_4h_50k (Tq=1,Tk=51200,H=4,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_4h_hd128 (Tq=1,Tk=256,H=4,D=128)", "status": "pass", "max_diff": 3.93e-06, "tolerance": 0.001, "c_time_us": 22.5, "pytorch_time_us": 101.4, "speedup": 4.51, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_4h_hd128 (Tq=1,Tk=256,H=4,D=128)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_gqa_16h_2kv (Tq=1,Tk=1024,H=16,D=64,kv=2)", "status": "pass", "max_diff": 3.55e-06, "tolerance": 0.001, "c_time_us": 94.1, "pytorch_time_us": 334.3, "speedup": 3.55, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_gqa_16h_2kv (Tq=1,Tk=1024,H=16,D=64,kv=2)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_mqa_16h_1kv (Tq=1,Tk=1024,H=16,D=64,kv=1)", "status": "pass", "max_diff": 3.28e-06, "tolerance": 0.001, "c_time_us": 97.2, "pytorch_time_us": 295.6, "speedup": 3.04, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_mqa_16h_1kv (Tq=1,Tk=1024,H=16,D=64,kv=1)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" }, { "name": "decode_32h_51200 (Tq=1,Tk=51200,H=32,D=64)", "status": "pass", "max_diff": 6.14e-07, "tolerance": 0.001, "c_time_us": 21081.3, "pytorch_time_us": 75812.9, "speedup": 3.6, "case_id": "Flash Attention (Causal, Online Softmax) - Performance (OMP Wrapper) :: tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1 :: decode_32h_51200 (Tq=1,Tk=51200,H=32,D=64)", "configuration": "tile_k_max=32, tile_heuristic=on, fast_exp=on, omp_threads=4, torch_threads=4, torch_interop=1", "evidence_kind": "numerical_and_performance" } ], "execution_kind": "make", "execution_id": "test-flash-attention", "execution_args": [] }, { "name": "FP16 Prefill/Decode Attention Contracts", "category": "kernels", "status": "pass", "duration_sec": 5.186627626419067, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "prefill_append_matches_decode_loop", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "prefill_append_matches_decode_loop", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_before(Q=5)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_before(Q=5)", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_after(Q=14)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_after(Q=14)", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_below_threshold(Q=63)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_below_threshold(Q=63)", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_at_threshold(Q=64)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_at_threshold(Q=64)", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128)", "configuration": "", "evidence_kind": "numerical" }, { "name": "unfused_f16_causal(T=17,H=4,D=32)", "status": "pass", "max_diff": 0.0, "tolerance": 3e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unfused_f16_causal(T=17,H=4,D=32)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_below_threshold(KV=256,S=256,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_below_threshold(KV=256,S=256,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_threshold(KV=257,S=512,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_threshold(KV=257,S=512,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_merge_rounding(KV=512,H=1,D=32,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_merge_rounding(KV=512,H=1,D=32,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_padded_gqa(KV=513,D=80,A=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_padded_gqa(KV=513,D=80,A=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_batch_store(KV=256,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_batch_store(KV=256,CAP=1034)", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_decode_append(KV=257,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_decode_append(KV=257,CAP=1034)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_below(KV=256,S=256)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_threshold(KV=257,S=512)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl(KV=1058)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_mid(KV=1307)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_long(KV=1609)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_single_range_prefill_route_qwen3vl(KV=1047)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_single_range_prefill_route_qwen3vl(KV=1047)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_fp32_preserves_legacy_abi(KV=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_fp32_preserves_legacy_abi(KV=256)", "configuration": "", "evidence_kind": "numerical" }, { "name": "unsupported_contract_rejected_without_output_write", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unsupported_contract_rejected_without_output_write", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_below(KV=256,S=256)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_threshold(KV=257,S=512)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl(KV=1058)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step20(KV=1047,P=1280)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step60(KV=1087,P=1280)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_mid(KV=1307)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_long(KV=1609)", "configuration": "", "evidence_kind": "numerical" }, { "name": "reject_single_fp32_reduction(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "reject_single_fp32_reduction(KV=1058)", "configuration": "", "evidence_kind": "numerical" }, { "name": "prefill_append_matches_decode_loop", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "prefill_append_matches_decode_loop :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_before(Q=5)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_before(Q=5) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_after(Q=14)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_after(Q=14) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_below_threshold(Q=63)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_below_threshold(Q=63) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_at_threshold(Q=64)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_at_threshold(Q=64) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "unfused_f16_causal(T=17,H=4,D=32)", "status": "pass", "max_diff": 0.0, "tolerance": 3e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unfused_f16_causal(T=17,H=4,D=32) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_below_threshold(KV=256,S=256,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_below_threshold(KV=256,S=256,C=1) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_threshold(KV=257,S=512,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_threshold(KV=257,S=512,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_merge_rounding(KV=512,H=1,D=32,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_merge_rounding(KV=512,H=1,D=32,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_padded_gqa(KV=513,D=80,A=128,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_padded_gqa(KV=513,D=80,A=128,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_batch_store(KV=256,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_batch_store(KV=256,CAP=1034) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_decode_append(KV=257,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_decode_append(KV=257,CAP=1034) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=16)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=16)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_below(KV=256,S=256) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_threshold(KV=257,S=512) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl(KV=1058) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_mid(KV=1307) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_long(KV=1609) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_single_range_prefill_route_qwen3vl(KV=1047)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_single_range_prefill_route_qwen3vl(KV=1047) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_fp32_preserves_legacy_abi(KV=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_fp32_preserves_legacy_abi(KV=256) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "unsupported_contract_rejected_without_output_write", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unsupported_contract_rejected_without_output_write :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_below(KV=256,S=256) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_threshold(KV=257,S=512) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl(KV=1058) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step20(KV=1047,P=1280) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step60(KV=1087,P=1280) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_mid(KV=1307) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_long(KV=1609) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "reject_single_fp32_reduction(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "reject_single_fp32_reduction(KV=1058) :: occurrence 2", "configuration": "", "evidence_kind": "numerical" }, { "name": "prefill_append_matches_decode_loop", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "prefill_append_matches_decode_loop :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_before(Q=5)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_before(Q=5) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_after(Q=14)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_after(Q=14) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_below_threshold(Q=63)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_below_threshold(Q=63) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_at_threshold(Q=64)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_at_threshold(Q=64) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "unfused_f16_causal(T=17,H=4,D=32)", "status": "pass", "max_diff": 0.0, "tolerance": 3e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unfused_f16_causal(T=17,H=4,D=32) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_below_threshold(KV=256,S=256,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_below_threshold(KV=256,S=256,C=1) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_threshold(KV=257,S=512,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_threshold(KV=257,S=512,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_merge_rounding(KV=512,H=1,D=32,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_merge_rounding(KV=512,H=1,D=32,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_padded_gqa(KV=513,D=80,A=128,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_padded_gqa(KV=513,D=80,A=128,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_batch_store(KV=256,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_batch_store(KV=256,CAP=1034) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_decode_append(KV=257,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_decode_append(KV=257,CAP=1034) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=20)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=20)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_below(KV=256,S=256) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_threshold(KV=257,S=512) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl(KV=1058) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_mid(KV=1307) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_long(KV=1609) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_single_range_prefill_route_qwen3vl(KV=1047)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_single_range_prefill_route_qwen3vl(KV=1047) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_fp32_preserves_legacy_abi(KV=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_fp32_preserves_legacy_abi(KV=256) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "unsupported_contract_rejected_without_output_write", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unsupported_contract_rejected_without_output_write :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_below(KV=256,S=256) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_threshold(KV=257,S=512) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl(KV=1058) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step20(KV=1047,P=1280) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step60(KV=1087,P=1280) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_mid(KV=1307) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_long(KV=1609) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "reject_single_fp32_reduction(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "reject_single_fp32_reduction(KV=1058) :: occurrence 3", "configuration": "", "evidence_kind": "numerical" }, { "name": "prefill_append_matches_decode_loop", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "prefill_append_matches_decode_loop :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_before(Q=5)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_before(Q=5) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_after(Q=14)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_after(Q=14) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_below_threshold(Q=63)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_below_threshold(Q=63) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_at_threshold(Q=64)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_at_threshold(Q=64) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "unfused_f16_causal(T=17,H=4,D=32)", "status": "pass", "max_diff": 0.0, "tolerance": 3e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unfused_f16_causal(T=17,H=4,D=32) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_below_threshold(KV=256,S=256,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_below_threshold(KV=256,S=256,C=1) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_threshold(KV=257,S=512,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_threshold(KV=257,S=512,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_merge_rounding(KV=512,H=1,D=32,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_merge_rounding(KV=512,H=1,D=32,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_padded_gqa(KV=513,D=80,A=128,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_padded_gqa(KV=513,D=80,A=128,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_batch_store(KV=256,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_batch_store(KV=256,CAP=1034) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_decode_append(KV=257,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_decode_append(KV=257,CAP=1034) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=24)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=24)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_below(KV=256,S=256) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_threshold(KV=257,S=512) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl(KV=1058) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_mid(KV=1307) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_long(KV=1609) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_single_range_prefill_route_qwen3vl(KV=1047)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_single_range_prefill_route_qwen3vl(KV=1047) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_fp32_preserves_legacy_abi(KV=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_fp32_preserves_legacy_abi(KV=256) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "unsupported_contract_rejected_without_output_write", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unsupported_contract_rejected_without_output_write :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_below(KV=256,S=256) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_threshold(KV=257,S=512) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl(KV=1058) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step20(KV=1047,P=1280) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step60(KV=1087,P=1280) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_mid(KV=1307) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_long(KV=1609) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" }, { "name": "reject_single_fp32_reduction(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "reject_single_fp32_reduction(KV=1058) :: occurrence 4", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "make", "execution_id": "test-attention-f16-split-kv", "execution_args": [] }, { "name": "Q4/Q6 Production Graph Parity", "category": "parity", "status": "pass", "duration_sec": 18.7550847530365, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-q4q6-llama-production-native", "execution_args": [] }, { "name": "Q4_K x Q8_K vs llama.cpp Performance", "category": "bench", "status": "pass", "duration_sec": 3.9396636486053467, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-q4k-q8k-llama-performance", "execution_args": [] }, { "name": "Qwen3.6 Q6_K M4 Prefill Performance", "category": "bench", "status": "pass", "duration_sec": 143.5380802154541, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-qwen36-q6k-m4-performance", "execution_args": [] }, { "name": "Whisper AVX2 FP16 GEMM Performance", "category": "bench", "status": "pass", "duration_sec": 1.426384687423706, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-whisper-f16-gemm-performance", "execution_args": [] }, { "name": "Numerical Replay Fixture Contract", "category": "parity", "status": "pass", "duration_sec": 0.48987293243408203, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-private-numerical-replay-runner", "execution_args": [] }, { "name": "v8 Stitched Dump Alignment", "category": "parity", "status": "pass", "duration_sec": 0.5631558895111084, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-dump-alignment", "execution_args": [] }, { "name": "Quantized Linear Scalar/Production/Threadpool Parity", "category": "parity", "status": "pass", "duration_sec": 52.645018577575684, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-threadpool-parity", "execution_args": [] }, { "name": "v7 Kernel Map Contracts", "category": "parity", "status": "pass", "duration_sec": 0.8160140514373779, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v7-kernel-map-contracts", "execution_args": [] }, { "name": "v7 Training Family Regression (full)", "category": "training", "status": "pass", "duration_sec": 277.67271995544434, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "regression-training-full", "execution_args": [] }, { "name": "v8 FP32 Generated Training Certification", "category": "training", "status": "pass", "duration_sec": 11.397255659103394, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v8-training-certify-fp32", "execution_args": [] }, { "name": "v8 FP32 Training-to-Inference Workflow", "category": "training", "status": "pass", "duration_sec": 118.64703273773193, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v8-training-workflow-fp32", "execution_args": [] }, { "name": "v8 FP32 BPE/GQA Training Matrix", "category": "training", "status": "pass", "duration_sec": 92.34770512580872, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v8-training-matrix-nightly", "execution_args": [] }, { "name": "v7 Backprop Long-Epoch Drift", "category": "training", "status": "pass", "duration_sec": 25.53773021697998, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v7-backprop-long-epoch-nightly", "execution_args": [] }, { "name": "v7 Backprop Kernel Parity", "category": "training", "status": "pass", "duration_sec": 3.360187530517578, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v7-kernel-parity-train", "execution_args": [] }, { "name": "v7 IR Visualizer E2E (nightly)", "category": "parity", "status": "pass", "duration_sec": 58.82748341560364, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v7-ir-visualizer-e2e-nightly", "execution_args": [] }, { "name": "v7 Visualizer Health (L1+L2)", "category": "parity", "status": "pass", "duration_sec": 0.8509020805358887, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "ir_visualizer_source", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "ir_visualizer_source", "configuration": "", "evidence_kind": "numerical" }, { "name": "dataset_viewer_source", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "dataset_viewer_source", "configuration": "", "evidence_kind": "numerical" }, { "name": "ir_hub_source", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "ir_hub_source", "configuration": "", "evidence_kind": "numerical" }, { "name": "L2_dataset_viewer", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "L2_dataset_viewer", "configuration": "", "evidence_kind": "numerical" }, { "name": "L2_ir_visualizer", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "L2_ir_visualizer", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "make", "execution_id": "v7-visualizer-health", "execution_args": [] }, { "name": "v7 Visualizer Generated E2E (L3)", "category": "parity", "status": "pass", "duration_sec": 1.7066574096679688, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "Generate_ir_report.html", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Generate_ir_report.html", "configuration": "", "evidence_kind": "numerical" }, { "name": "Generate_dataset_viewer.html", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Generate_dataset_viewer.html", "configuration": "", "evidence_kind": "numerical" }, { "name": "Generate_ir_hub.html", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Generate_ir_hub.html", "configuration": "", "evidence_kind": "numerical" }, { "name": "L1_Health__ir_report.html", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "L1_Health__ir_report.html", "configuration": "", "evidence_kind": "numerical" }, { "name": "L1_Health__ir_hub.html", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "L1_Health__ir_hub.html", "configuration": "", "evidence_kind": "numerical" }, { "name": "Embedded_JSON__ir_report.html", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Embedded_JSON__ir_report.html", "configuration": "", "evidence_kind": "numerical" }, { "name": "Structure__ir_hub.html", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Structure__ir_hub.html", "configuration": "", "evidence_kind": "numerical" }, { "name": "Cross-artifact_Consistency", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "Cross-artifact_Consistency", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "make", "execution_id": "v7-visualizer-generated-e2e", "execution_args": [] }, { "name": "v7 Core Stabilization Matrix", "category": "training", "status": "pass", "duration_sec": 526.1413805484772, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v7-stabilization-nightly", "execution_args": [] }, { "name": "v8 Inference Runtime Contracts", "category": "inference", "status": "pass", "duration_sec": 1.3458960056304932, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "ir_visualizer_source", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "ir_visualizer_source", "configuration": "", "evidence_kind": "numerical" }, { "name": "dataset_viewer_source", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "dataset_viewer_source", "configuration": "", "evidence_kind": "numerical" }, { "name": "ir_hub_source", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "ir_hub_source", "configuration": "", "evidence_kind": "numerical" }, { "name": "L2_dataset_viewer", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "L2_dataset_viewer", "configuration": "", "evidence_kind": "numerical" }, { "name": "L2_ir_visualizer", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "L2_ir_visualizer", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "make", "execution_id": "v8-validate-contracts", "execution_args": [] }, { "name": "v8 Kernel Map Contracts", "category": "inference", "status": "pass", "duration_sec": 34.53775453567505, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v8-kernel-map-contracts", "execution_args": [] }, { "name": "v8 Model Memory Plan Matrix", "category": "inference", "status": "pass", "duration_sec": 118.56707096099854, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v8-model-memory-plans-nightly", "execution_args": [] }, { "name": "v8 Numerical Kernel Contracts", "category": "parity", "status": "skip", "duration_sec": 66.36240196228027, "stdout": "", "stderr": "", "error_msg": "SKIP: native AVX-512 BF16 or AMX-BF16 hardware is required", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "rmsnorm_fp64_sum_3x128", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "rmsnorm_fp64_sum_3x128", "configuration": "", "evidence_kind": "numerical" }, { "name": "rmsnorm_fp64_sum_2x4096", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "rmsnorm_fp64_sum_2x4096", "configuration": "", "evidence_kind": "numerical" }, { "name": "qk_norm_fp64_sum_decode", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qk_norm_fp64_sum_decode", "configuration": "", "evidence_kind": "numerical" }, { "name": "prefill_append_matches_decode_loop", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "prefill_append_matches_decode_loop", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_before(Q=5)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_before(Q=5)", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_short_text_after(Q=14)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_short_text_after(Q=14)", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_below_threshold(Q=63)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_below_threshold(Q=63)", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_at_threshold(Q=64)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_at_threshold(Q=64)", "configuration": "", "evidence_kind": "numerical" }, { "name": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "flash_auto_qwen3vl_visual(Q=1008,KV=1013,H=32,Hkv=8,D=128)", "configuration": "", "evidence_kind": "numerical" }, { "name": "unfused_f16_causal(T=17,H=4,D=32)", "status": "pass", "max_diff": 0.0, "tolerance": 3e-05, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unfused_f16_causal(T=17,H=4,D=32)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_below_threshold(KV=256,S=256,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_below_threshold(KV=256,S=256,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_threshold(KV=257,S=512,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_threshold(KV=257,S=512,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_merge_rounding(KV=512,H=1,D=32,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_merge_rounding(KV=512,H=1,D=32,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl(KV=1058,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step20(KV=1047,P=1280,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_step60(KV=1087,P=1280,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_mid(KV=1307,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_qwen3vl_long(KV=1609,H=32,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "f16_split_padded_gqa(KV=513,D=80,A=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "f16_split_padded_gqa(KV=513,D=80,A=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_capacity_independent(KV=257,CAP=1034,S=512,H=8,Hkv=2,D=256,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_batch_store(KV=256,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_batch_store(KV=256,CAP=1034)", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen35_f16_cache_decode_append(KV=257,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen35_f16_cache_decode_append(KV=257,CAP=1034)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen35_cache_write_through(KV=257,CAP=1034,S=512,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_short_decode(KV=21,CAP=1034,P=1280,H=32,Hkv=2,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "glm4_f16_cache_prefill_then_decode(KV=33,CAP=1034)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=1)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_glm4_cache_write_through(KV=33,P=1280,H=32,Hkv=2,D=128,C=1)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_below(KV=256,S=256)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_threshold(KV=257,S=512)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl(KV=1058)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step20(KV=1047,P=1280)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_step60(KV=1087,P=1280)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_mid(KV=1307)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_contract_route_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_contract_route_qwen3vl_long(KV=1609)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_single_range_prefill_route_qwen3vl(KV=1047)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_single_range_prefill_route_qwen3vl(KV=1047)", "configuration": "", "evidence_kind": "numerical" }, { "name": "explicit_fp32_preserves_legacy_abi(KV=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "explicit_fp32_preserves_legacy_abi(KV=256)", "configuration": "", "evidence_kind": "numerical" }, { "name": "unsupported_contract_rejected_without_output_write", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "unsupported_contract_rejected_without_output_write", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_below(KV=256,S=256)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_below(KV=256,S=256)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_threshold(KV=257,S=512)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_threshold(KV=257,S=512)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl(KV=1058)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step20(KV=1047,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step20(KV=1047,P=1280)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_step60(KV=1087,P=1280)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_step60(KV=1087,P=1280)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_mid(KV=1307)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_mid(KV=1307)", "configuration": "", "evidence_kind": "numerical" }, { "name": "llama_oracle_qwen3vl_long(KV=1609)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "llama_oracle_qwen3vl_long(KV=1609)", "configuration": "", "evidence_kind": "numerical" }, { "name": "reject_single_fp32_reduction(KV=1058)", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "reject_single_fp32_reduction(KV=1058)", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "make", "execution_id": "test-numerical-contracts", "execution_args": [] }, { "name": "Cohere/Laguna Compiler/Circuit Contracts", "category": "inference", "status": "pass", "duration_sec": 6.270699739456177, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-cohere-laguna-contracts", "execution_args": [] }, { "name": "v8 Real-Manifest Lowering/Compile Matrix", "category": "inference", "status": "pass", "duration_sec": 12.641839265823364, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-artifact-compile-matrix", "execution_args": [] }, { "name": "Muse-Glimmer Text Compiler/Kernel Contracts", "category": "parity", "status": "pass", "duration_sec": 11.470269203186035, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-muse-glimmer-contracts", "execution_args": [] }, { "name": "v8 Capability Case Coverage", "category": "inference", "status": "pass", "duration_sec": 1.0580239295959473, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-capability-cases", "execution_args": [] }, { "name": "Qwen3.6/3.8 Dense Metadata and Chat Contracts", "category": "inference", "status": "pass", "duration_sec": 1.2193140983581543, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-qwen38-dense-contracts", "execution_args": [] }, { "name": "Qwen3.5 Routed-MoE Storage and Execution Contracts", "category": "parity", "status": "pass", "duration_sec": 1.3328118324279785, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-routed-moe-contracts", "execution_args": [] }, { "name": "Qwen3.8 Flash Circuit and Numerical Providers", "category": "parity", "status": "pass", "duration_sec": 8.630710363388062, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-qwen38-flash-contracts", "execution_args": [] }, { "name": "Command A+ Native NVFP4 / X-ray Contracts", "category": "parity", "status": "pass", "duration_sec": 3.2247121334075928, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-command-a-plus-nvfp4", "execution_args": [] }, { "name": "v8 Numerical X-ray Contracts", "category": "bf16", "status": "pass", "duration_sec": 11.71196436882019, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-bf16-xray", "execution_args": [] }, { "name": "v8 X-ray Injected-Fault Validator", "category": "bf16", "status": "pass", "duration_sec": 0.5542263984680176, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-xray-validator-selftest", "execution_args": [] }, { "name": "BF16 PyTorch/oneDNN Production-Shape Oracle", "category": "bf16", "status": "skip", "duration_sec": 0.32779598236083984, "stdout": "", "stderr": "", "error_msg": "SKIP: exact oneDNN 3.7.1 source/build unavailable", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-bf16-pytorch-onednn-oracle-auto", "execution_args": [] }, { "name": "Qwen3-VL BF16 Generated Kernel Oracle Matrix", "category": "bf16", "status": "skip", "duration_sec": 24.603260040283203, "stdout": "", "stderr": "", "error_msg": "SKIP: unittest/bf16/test_patch_projection_pytorch_onednn_bf16.py requires a USE_ONEDNN=1 engine", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-qwen3vl-bf16-kernel-oracles", "execution_args": [] }, { "name": "Qwen3-VL BF16 Private 40-Image E2E Parity", "category": "bf16", "status": "skip", "duration_sec": 0.34146714210510254, "stdout": "", "stderr": "", "error_msg": "Private corpus lane is not configured on this runner.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-qwen3vl-bf16-private-corpus-parity-auto", "execution_args": [] }, { "name": "v8 DSL Zero-Hardcoding Policy", "category": "inference", "status": "pass", "duration_sec": 16.571309328079224, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-dsl-policy", "execution_args": [] }, { "name": "Repository Source Integrity", "category": "inference", "status": "pass", "duration_sec": 2.867616891860962, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-source-integrity", "execution_args": [] }, { "name": "v8 Template Circuit/Dataflow Audit", "category": "inference", "status": "pass", "duration_sec": 0.438701868057251, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-template-circuit-audit", "execution_args": [] }, { "name": "Architecture Contract Dashboard", "category": "inference", "status": "pass", "duration_sec": 35.76387929916382, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "ir_visualizer_source", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "ir_visualizer_source", "configuration": "", "evidence_kind": "numerical" }, { "name": "dataset_viewer_source", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "dataset_viewer_source", "configuration": "", "evidence_kind": "numerical" }, { "name": "ir_hub_source", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "ir_hub_source", "configuration": "", "evidence_kind": "numerical" }, { "name": "L2_dataset_viewer", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "L2_dataset_viewer", "configuration": "", "evidence_kind": "numerical" }, { "name": "L2_ir_visualizer", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "L2_ir_visualizer", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "make", "execution_id": "test-architecture-contracts", "execution_args": [] }, { "name": "v8 Inference Regression (fast)", "category": "inference", "status": "pass", "duration_sec": 470.62741923332214, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v8-regression-fast", "execution_args": [] }, { "name": "v8 Vision Visualizer Artifacts", "category": "inference", "status": "pass", "duration_sec": 0.6902246475219727, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "derived_payload", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "derived_payload", "configuration": "", "evidence_kind": "numerical" }, { "name": "html_embedding", "status": "pass", "max_diff": 0.0, "tolerance": 1.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "html_embedding", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "make", "execution_id": "v8-visualizer-vision-artifacts", "execution_args": [] }, { "name": "v8 Vision Kernel Parity", "category": "kernels", "status": "pass", "duration_sec": 6.054496765136719, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "im2patch (Image 224x224, Patch 16)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 421.0, "pytorch_time_us": 1014.0, "speedup": 2.4085510688836105, "case_id": "im2patch (Image 224x224, Patch 16)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "patch2im (Image 224x224, Patch 16)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 104.0, "pytorch_time_us": 810.0, "speedup": 7.788461538461539, "case_id": "patch2im (Image 224x224, Patch 16)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "position_embeddings_add (2304 tokens, dim 1152)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 1911.0, "pytorch_time_us": 1701.0, "speedup": 0.8901098901098902, "case_id": "position_embeddings_add (2304 tokens, dim 1152)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "position_embeddings_add_tiled_2d (6x6, dim 8, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 75.0, "pytorch_time_us": null, "speedup": null, "case_id": "position_embeddings_add_tiled_2d (6x6, dim 8, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "position_embeddings_add_tiled_2d Qwen3-VL resize (56x72 from 48x48, dim 1152, merge 2)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "position_embeddings_add_tiled_2d Qwen3-VL resize (56x72 from 48x48, dim 1152, merge 2)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "position_embeddings_add_tiled_2d ragged edges (5x7, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "position_embeddings_add_tiled_2d ragged edges (5x7, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "vision_position_ids_2d_merge (4x4, merge 2)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "vision_position_ids_2d_merge (4x4, merge 2)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "vision_position_ids_2d_merge (6x6, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "vision_position_ids_2d_merge (6x6, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "rowwise_bias_add (2304 tokens, dim 1152)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 1431.0, "pytorch_time_us": 1165.0, "speedup": 0.8141160027952481, "case_id": "rowwise_bias_add (2304 tokens, dim 1152)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "add_stream_reorder_2d (6x6, dim 8, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 103.0, "pytorch_time_us": null, "speedup": null, "case_id": "add_stream_reorder_2d (6x6, dim 8, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "spatial_merge_2x2 (48x48, dim 64)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 268.0, "pytorch_time_us": 187.0, "speedup": 0.6977611940298507, "case_id": "spatial_merge_2x2 (48x48, dim 64)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "spatial_merge_contiguous_tiled (6x6, dim 5, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 32.0, "pytorch_time_us": null, "speedup": null, "case_id": "spatial_merge_contiguous_tiled (6x6, dim 5, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "feature_slice_copy (6 rows, src 4, dst 11, off 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 35.0, "pytorch_time_us": null, "speedup": null, "case_id": "feature_slice_copy (6 rows, src 4, dst 11, off 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "feature_concat (5 rows, main 3, branch 2 x 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 77.0, "pytorch_time_us": null, "speedup": null, "case_id": "feature_concat (5 rows, main 3, branch 2 x 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "feature_concat in-place expand (5 rows, main 3, branch 2 x 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "feature_concat in-place expand (5 rows, main 3, branch 2 x 3)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 8) Q", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 8) Q", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 8) K", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 8) K", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 72) Q", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 72) Q", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 72) K", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 72) K", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "im2patch (Image 32x32, Patch 8)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 45.0, "pytorch_time_us": 322.0, "speedup": 7.155555555555556, "case_id": "im2patch (Image 32x32, Patch 8)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "patch2im (Image 32x32, Patch 8)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 38.0, "pytorch_time_us": 88.0, "speedup": 2.3157894736842106, "case_id": "patch2im (Image 32x32, Patch 8)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" } ], "execution_kind": "make", "execution_id": "test-v8-vision-kernels", "execution_args": [] }, { "name": "Qwen3-VL Circuit/Codegen Contracts", "category": "inference", "status": "pass", "duration_sec": 21.45914316177368, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-qwen3vl", "execution_args": [] }, { "name": "Qwen3-VL Methodical Layer Parity", "category": "parity", "status": "pass", "duration_sec": 29.915873289108276, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [ { "name": "qwen3vl_checkpoint_coverage", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen3vl_checkpoint_coverage", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen3vl_circuit_codegen", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen3vl_circuit_codegen", "configuration": "", "evidence_kind": "numerical" }, { "name": "im2patch (Image 224x224, Patch 16)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 460.0, "pytorch_time_us": 1029.0, "speedup": 2.2369565217391303, "case_id": "im2patch (Image 224x224, Patch 16)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "patch2im (Image 224x224, Patch 16)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 101.0, "pytorch_time_us": 868.0, "speedup": 8.594059405940593, "case_id": "patch2im (Image 224x224, Patch 16)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "position_embeddings_add (2304 tokens, dim 1152)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 2047.0000000000002, "pytorch_time_us": 1179.0, "speedup": 0.5759648265754763, "case_id": "position_embeddings_add (2304 tokens, dim 1152)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "position_embeddings_add_tiled_2d (6x6, dim 8, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 107.0, "pytorch_time_us": null, "speedup": null, "case_id": "position_embeddings_add_tiled_2d (6x6, dim 8, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "position_embeddings_add_tiled_2d Qwen3-VL resize (56x72 from 48x48, dim 1152, merge 2)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "position_embeddings_add_tiled_2d Qwen3-VL resize (56x72 from 48x48, dim 1152, merge 2)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "position_embeddings_add_tiled_2d ragged edges (5x7, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "position_embeddings_add_tiled_2d ragged edges (5x7, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "vision_position_ids_2d_merge (4x4, merge 2)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "vision_position_ids_2d_merge (4x4, merge 2)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "vision_position_ids_2d_merge (6x6, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "vision_position_ids_2d_merge (6x6, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "rowwise_bias_add (2304 tokens, dim 1152)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 1485.0, "pytorch_time_us": 1256.0, "speedup": 0.8457912457912458, "case_id": "rowwise_bias_add (2304 tokens, dim 1152)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "add_stream_reorder_2d (6x6, dim 8, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 84.0, "pytorch_time_us": null, "speedup": null, "case_id": "add_stream_reorder_2d (6x6, dim 8, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "spatial_merge_2x2 (48x48, dim 64)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 295.0, "pytorch_time_us": 188.0, "speedup": 0.6372881355932204, "case_id": "spatial_merge_2x2 (48x48, dim 64)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "spatial_merge_contiguous_tiled (6x6, dim 5, merge 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 33.0, "pytorch_time_us": null, "speedup": null, "case_id": "spatial_merge_contiguous_tiled (6x6, dim 5, merge 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "feature_slice_copy (6 rows, src 4, dst 11, off 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 35.0, "pytorch_time_us": null, "speedup": null, "case_id": "feature_slice_copy (6 rows, src 4, dst 11, off 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "feature_concat (5 rows, main 3, branch 2 x 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 78.0, "pytorch_time_us": null, "speedup": null, "case_id": "feature_concat (5 rows, main 3, branch 2 x 3)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "feature_concat in-place expand (5 rows, main 3, branch 2 x 3)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "feature_concat in-place expand (5 rows, main 3, branch 2 x 3)", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 8) Q", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 8) Q", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 8) K", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 8) K", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 72) Q", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 72) Q", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "mrope_qk_vision (2 heads, 4 tokens, dim 72) K", "status": "not_tested", "max_diff": 2.38e-07, "tolerance": null, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "mrope_qk_vision (2 heads, 4 tokens, dim 72) K", "configuration": "", "evidence_kind": "numerical_measurement" }, { "name": "im2patch (Image 32x32, Patch 8)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 46.0, "pytorch_time_us": 330.0, "speedup": 7.173913043478262, "case_id": "im2patch (Image 32x32, Patch 8)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "patch2im (Image 32x32, Patch 8)", "status": "not_tested", "max_diff": 0.0, "tolerance": null, "c_time_us": 36.0, "pytorch_time_us": 86.0, "speedup": 2.388888888888889, "case_id": "patch2im (Image 32x32, Patch 8)", "configuration": "", "evidence_kind": "numerical_measurement_and_performance" }, { "name": "qwen3vl_frontend_mrope", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen3vl_frontend_mrope", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen3vl_attention_contract", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen3vl_attention_contract", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen3vl_q8_projection_matrix", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen3vl_q8_projection_matrix", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen3vl_eos_contract", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen3vl_eos_contract", "configuration": "", "evidence_kind": "numerical" }, { "name": "qwen3vl_segmented_dump_alignment", "status": "pass", "max_diff": 0.0, "tolerance": 0.0, "c_time_us": null, "pytorch_time_us": null, "speedup": null, "case_id": "qwen3vl_segmented_dump_alignment", "configuration": "", "evidence_kind": "numerical" } ], "execution_kind": "make", "execution_id": "test-qwen3vl-methodical-parity", "execution_args": [] }, { "name": "Qwen3-VL Private 40-Image E2E Parity", "category": "parity", "status": "skip", "duration_sec": 0.3363308906555176, "stdout": "", "stderr": "", "error_msg": "Private corpus lane is not configured on this runner.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-qwen3vl-private-corpus-parity-auto", "execution_args": [] }, { "name": "Qwen3.6-VL Private 40-Image E2E Parity", "category": "parity", "status": "skip", "duration_sec": 0.3442823886871338, "stdout": "", "stderr": "", "error_msg": "Private corpus lane is not configured on this runner.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-qwen36vl-private-corpus-parity-auto", "execution_args": [] }, { "name": "Gemma4 Private 40-Image First-Token Parity", "category": "parity", "status": "skip", "duration_sec": 0.35512399673461914, "stdout": "", "stderr": "", "error_msg": "Private corpus lane is not configured on this runner.", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-gemma4-private-corpus-parity-auto", "execution_args": [] }, { "name": "v8 Qwen3-VL Vision Smoke", "category": "inference", "status": "skip", "duration_sec": 0.3505058288574219, "stdout": "", "stderr": "", "error_msg": "SKIP: Qwen3-VL cached decoder/mmproj not found under /opt/app-root/src/.cache/ck-engine-v8/models/Qwen--Qwen3-VL-8B-Instruct-GGUF", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-qwen3vl-e2e-smoke", "execution_args": [] }, { "name": "v8 Qwen3-VL Q8 Encoder Accuracy (llama.cpp artifact gate)", "category": "parity", "status": "pass", "duration_sec": 225.81748914718628, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "vision-encoder-full", "execution_args": [] }, { "name": "v8 Gemma4 Vision Smoke", "category": "inference", "status": "skip", "duration_sec": 0.33816003799438477, "stdout": "", "stderr": "", "error_msg": "SKIP: Gemma4 vision smoke needs >=50 GiB MemAvailable; found 14 GiB", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-gemma4-vision-smoke", "execution_args": [] }, { "name": "v8 Decoder Matrix (quick)", "category": "inference", "status": "pass", "duration_sec": 0.3943490982055664, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-decoder-matrix-quick", "execution_args": [] }, { "name": "v8 Gemma4 High-Memory Certification", "category": "inference", "status": "skip", "duration_sec": 0.4154179096221924, "stdout": "", "stderr": "", "error_msg": "SKIP: Gemma4 certification needs 50 GiB MemAvailable; found 14.5 GiB", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-gemma4-highmem", "execution_args": [] }, { "name": "v8 Nemotron Nano 9B High-Memory Smoke", "category": "inference", "status": "skip", "duration_sec": 0.34822773933410645, "stdout": "", "stderr": "", "error_msg": "SKIP: Nemotron Nano 9B v2 v8 certification needs >=16 GiB MemAvailable; found 14 GiB", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-nemotron9-highmem", "execution_args": [] }, { "name": "v8 GLM-4 9B High-Memory Smoke", "category": "inference", "status": "skip", "duration_sec": 0.3384110927581787, "stdout": "", "stderr": "", "error_msg": "SKIP: GLM-4 9B v8 smoke needs >=70 GiB MemAvailable; found 14 GiB", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-glm4-highmem", "execution_args": [] }, { "name": "v8 Kimi High-Memory Smoke", "category": "inference", "status": "skip", "duration_sec": 0.3438096046447754, "stdout": "", "stderr": "", "error_msg": "SKIP: Kimi high-memory smoke requires V8_KIMI_MODEL=/path/to/model-or-hf-uri", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-kimi-highmem", "execution_args": [] }, { "name": "v8 Qwen3.6-27B High-Memory Coherent-Answer Smoke", "category": "inference", "status": "skip", "duration_sec": 0.34128713607788086, "stdout": "", "stderr": "", "error_msg": "SKIP: Qwen3.6 high-memory smoke requires V8_QWEN36_MODEL=/path/to/runtime-or-gguf", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-qwen36-highmem", "execution_args": [] }, { "name": "v8 Xeon Family Compiler/Circuit Contracts", "category": "inference", "status": "pass", "duration_sec": 73.03131318092346, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-xeon-family-contracts", "execution_args": [] }, { "name": "v8 Serve Localhost E2E (Qwen3-0.6B)", "category": "inference", "status": "pass", "duration_sec": 80.13019919395447, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "test-v8-serve-localhost-e2e", "execution_args": [] }, { "name": "GEMM Benchmark", "category": "bench", "status": "pass", "duration_sec": 63.6871919631958, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": 156.34, "perf_unit": "GFLOPS", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "bench_gemm", "execution_args": [] }, { "name": "Q4_K Dispatch Matrix", "category": "bench", "status": "pass", "duration_sec": 3.868767738342285, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "bench-q4k-dispatch-matrix-quick", "execution_args": [] }, { "name": "Q8_0 FP32 GEMM", "category": "bench", "status": "pass", "duration_sec": 0.8927731513977051, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": 0.999, "perf_unit": "speedup", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "bench-q8-0-fp32-gemm-quick", "execution_args": [] }, { "name": "Q4_K Gate/Up x16 Chunk4", "category": "bench", "status": "pass", "duration_sec": 0.7763848304748535, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": 0.779, "perf_unit": "speedup", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "bench-q4k-gateup-swiglu-x16-chunk4-quick", "execution_args": [] } ], "capability_evidence": { "schema": "cke.v8.capability_evidence_report", "schema_version": 1, "scope": "current_run", "source": { "repository_commit": "19694d9d9ba7cef51bce0e5ee7229041802bcb9e", "manifest_sha256": "f26ecacd7a4e18edefe738a5230fc391e3b87d754b5cc7fee8200ae7c9737c74", "event": "schedule" }, "summary": { "total": 12, "passed": 9, "failed": 0, "errors": 0, "timeouts": 0, "not_tested": 3 }, "cases": [ { "id": "cohere2.compiler-circuit", "family": "cohere2", "circuits": [ "cohere2" ], "evidence_level": "contract", "phases": [ "conversion", "lowering" ], "oracle": "none", "artifact": { "kind": "synthetic_fixture" }, "schedule": { "events": [ "pull_request", "nightly" ], "resource_class": "portable" }, "status": "pass", "selected": true, "executed": true, "reason": "", "execution": { "kind": "make", "id": "test-v8-cohere-laguna-contracts", "args": [], "name": "Cohere/Laguna Compiler/Circuit Contracts", "duration_sec": 6.270699739456177, "error": "" } }, { "id": "laguna.compiler-circuit", "family": "laguna", "circuits": [ "laguna" ], "evidence_level": "contract", "phases": [ "conversion", "lowering" ], "oracle": "none", "artifact": { "kind": "synthetic_fixture" }, "schedule": { "events": [ "pull_request", "nightly" ], "resource_class": "portable" }, "status": "pass", "selected": true, "executed": true, "reason": "", "execution": { "kind": "make", "id": "test-v8-cohere-laguna-contracts", "args": [], "name": "Cohere/Laguna Compiler/Circuit Contracts", "duration_sec": 6.270699739456177, "error": "" } }, { "id": "laguna-xs.real-manifest-compile", "family": "laguna", "circuits": [ "laguna" ], "evidence_level": "contract", "phases": [ "lowering", "compile", "prefill", "decode" ], "oracle": "none", "artifact": { "kind": "real_manifest", "catalog_id": "laguna_xs_2_1" }, "schedule": { "events": [ "pull_request", "nightly" ], "resource_class": "portable" }, "status": "pass", "selected": true, "executed": true, "reason": "", "execution": { "kind": "make", "id": "test-v8-artifact-compile-matrix", "args": [], "name": "v8 Real-Manifest Lowering/Compile Matrix", "duration_sec": 12.641839265823364, "error": "" } }, { "id": "nemotron-nano-9b-v2.real-manifest-compile", "family": "nemotron_h", "circuits": [ "nemotron_h" ], "evidence_level": "contract", "phases": [ "lowering", "compile", "prefill", "decode" ], "oracle": "none", "artifact": { "kind": "real_manifest", "catalog_id": "nemotron_nano_9b" }, "schedule": { "events": [ "pull_request", "nightly" ], "resource_class": "portable" }, "status": "pass", "selected": true, "executed": true, "reason": "", "execution": { "kind": "make", "id": "test-v8-artifact-compile-matrix", "args": [], "name": "v8 Real-Manifest Lowering/Compile Matrix", "duration_sec": 12.641839265823364, "error": "" } }, { "id": "qwen35-moe.storage-lowering", "family": "qwen35", "circuits": [ "qwen35" ], "evidence_level": "component_oracle", "phases": [ "conversion", "lowering", "prefill", "decode" ], "oracle": "scalar", "artifact": { "kind": "synthetic_fixture" }, "schedule": { "events": [ "pull_request", "nightly" ], "resource_class": "portable" }, "status": "pass", "selected": true, "executed": true, "reason": "", "execution": { "kind": "make", "id": "test-v8-routed-moe-contracts", "args": [], "name": "Qwen3.5 Routed-MoE Storage and Execution Contracts", "duration_sec": 1.3328118324279785, "error": "" } }, { "id": "qwen36-dense.storage-lowering", "family": "qwen36", "circuits": [ "qwen35" ], "evidence_level": "contract", "phases": [ "conversion", "lowering" ], "oracle": "none", "artifact": { "kind": "synthetic_fixture" }, "schedule": { "events": [ "pull_request", "nightly" ], "resource_class": "portable" }, "status": "pass", "selected": true, "executed": true, "reason": "", "execution": { "kind": "make", "id": "test-v8-qwen38-dense-contracts", "args": [], "name": "Qwen3.6/3.8 Dense Metadata and Chat Contracts", "duration_sec": 1.2193140983581543, "error": "" } }, { "id": "qwen38-dense.storage-lowering", "family": "qwen38", "circuits": [ "qwen38" ], "evidence_level": "contract", "phases": [ "conversion", "lowering" ], "oracle": "none", "artifact": { "kind": "synthetic_fixture" }, "schedule": { "events": [ "pull_request", "nightly" ], "resource_class": "portable" }, "status": "pass", "selected": true, "executed": true, "reason": "", "execution": { "kind": "make", "id": "test-v8-qwen38-dense-contracts", "args": [], "name": "Qwen3.6/3.8 Dense Metadata and Chat Contracts", "duration_sec": 1.2193140983581543, "error": "" } }, { "id": "qwen38-flash.provider-oracles", "family": "qwen38_flash", "circuits": [ "qwen4_exp" ], "evidence_level": "component_oracle", "phases": [ "prefill", "decode" ], "oracle": "scalar", "artifact": { "kind": "synthetic_fixture" }, "schedule": { "events": [ "pull_request", "nightly" ], "resource_class": "portable" }, "status": "pass", "selected": true, "executed": true, "reason": "", "execution": { "kind": "make", "id": "test-v8-qwen38-flash-contracts", "args": [], "name": "Qwen3.8 Flash Circuit and Numerical Providers", "duration_sec": 8.630710363388062, "error": "" } }, { "id": "command-a-plus.nvfp4-oracle", "family": "command_a_plus", "circuits": [ "cohere_command_a_plus_text" ], "evidence_level": "component_oracle", "phases": [ "lowering", "prefill", "decode" ], "oracle": "scalar", "artifact": { "kind": "synthetic_fixture" }, "schedule": { "events": [ "pull_request", "nightly" ], "resource_class": "portable" }, "status": "pass", "selected": true, "executed": true, "reason": "", "execution": { "kind": "make", "id": "test-v8-command-a-plus-nvfp4", "args": [], "name": "Command A+ Native NVFP4 / X-ray Contracts", "duration_sec": 3.2247121334075928, "error": "" } }, { "id": "laguna-s.long-context", "family": "laguna", "circuits": [ "laguna" ], "evidence_level": "full_artifact", "phases": [ "conversion", "lowering", "compile", "load", "prefill", "decode", "quality" ], "oracle": "llama.cpp", "artifact": { "kind": "real_model", "catalog_id": "laguna_s_2_1" }, "schedule": { "events": [ "hardware" ], "resource_class": "high_memory" }, "status": "not_tested", "selected": false, "executed": false, "reason": "entry point was not selected for this run", "execution": null }, { "id": "cohere2-command-r7b.long-context", "family": "cohere2", "circuits": [ "cohere2" ], "evidence_level": "full_artifact", "phases": [ "conversion", "lowering", "compile", "load", "prefill", "decode", "quality" ], "oracle": "llama.cpp", "artifact": { "kind": "real_model", "catalog_id": "cohere2_command_r7b" }, "schedule": { "events": [ "hardware" ], "resource_class": "high_memory" }, "status": "not_tested", "selected": false, "executed": false, "reason": "entry point was not selected for this run", "execution": null }, { "id": "qwen38-dense.long-context", "family": "qwen38", "circuits": [ "qwen38" ], "evidence_level": "full_artifact", "phases": [ "conversion", "lowering", "compile", "load", "prefill", "decode", "quality" ], "oracle": "llama.cpp", "artifact": { "kind": "real_model", "catalog_id": "qwen38_27b" }, "schedule": { "events": [ "hardware" ], "resource_class": "high_memory" }, "status": "not_tested", "selected": false, "executed": false, "reason": "entry point was not selected for this run", "execution": null } ], "errors": [] }, "architecture_contracts": { "families": [ { "family": "qwen2", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "fast v8 regression family", "runtime": "pass", "template": "pass" }, { "family": "qwen3", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "QK-norm + GGUF smoke", "runtime": "pass", "template": "pass" }, { "family": "qwen3.5", "generated_c": "pass", "ir": "pass", "model": "partial", "notes": "long-generation numerical parity monitored", "runtime": "partial", "template": "pass" }, { "family": "qwen3-vl", "generated_c": "partial", "ir": "pass", "model": "pass", "notes": "promoted vision smoke; bridge parity still active", "runtime": "partial", "template": "pass" }, { "family": "gemma3", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "split-half RoPE and sliding attention covered", "runtime": "pass", "template": "pass" }, { "family": "gemma4", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "text coherent; vision bridge early", "runtime": "partial", "template": "pass" }, { "family": "glm4", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "partial pairwise RoPE + GGUF smoke", "runtime": "partial", "template": "pass" }, { "family": "nemotron-h", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "Mamba2 path equivalence monitored", "runtime": "partial", "template": "pass" }, { "family": "kimi-vl", "generated_c": "partial", "ir": "partial", "model": "partial", "notes": "MLA/MoE scalar contracts added; full template lowering/tokenizer/vision bridge in bring-up", "runtime": "partial", "template": "pass" }, { "family": "llama/nanbeige", "generated_c": "pass", "ir": "pass", "model": "partial", "notes": "Nanbeige active bring-up lane", "runtime": "partial", "template": "pass" } ], "generated_at": "2026-09-21T08:39:02.186572Z", "sections": [ { "checks_failed": 0, "checks_passed": 12, "details": "56 explicit critical edges; 69 implicit/missing edges across promoted templates.", "id": "template_circuit", "label": "Template / Circuit", "rows": [ { "explicit_count": 0, "missing_count": 5, "status": "warn", "template": "qwen2", "warnings": [ "decoder.footer[2].logits.x", "decoder.body[1].qkv_proj.x", "decoder.body[4].out_proj.x", "decoder.body[7].mlp_gate_up.x", "decoder.body[9].mlp_down.x" ] }, { "explicit_count": 0, "missing_count": 5, "status": "warn", "template": "qwen3", "warnings": [ "decoder.footer[2].logits.x", "decoder.body[1].qkv_proj.x", "decoder.body[5].out_proj.x", "decoder.body[8].mlp_gate_up.x", "decoder.body[10].mlp_down.x" ] }, { "explicit_count": 12, "missing_count": 14, "status": "warn", "template": "qwen35", "warnings": [ "decoder.footer[2].logits.x", "decoder.body:recurrent[1].recurrent_qkv_proj.x", "decoder.body:recurrent[2].recurrent_gate_proj.x", "decoder.body:recurrent[3].recurrent_alpha_proj.x", "decoder.body:recurrent[4].recurrent_beta_proj.x", "decoder.body:recurrent[14].recurrent_out_proj.x", "decoder.body:recurrent[18].mlp_gate_up.x", "decoder.body:recurrent[20].mlp_down.x" ] }, { "explicit_count": 0, "missing_count": 5, "status": "warn", "template": "qwen3vl", "warnings": [ "decoder.footer[2].logits.x", "decoder.body[1].qkv_proj.x", "decoder.body[5].out_proj.x", "decoder.body[8].mlp_gate_up.x", "decoder.body[10].mlp_down.x" ] }, { "explicit_count": 0, "missing_count": 3, "status": "warn", "template": "qwen3_vl_vision", "warnings": [ "vision_encoder.body[5].out_proj.x", "vision_encoder.body[8].mlp_up.x", "vision_encoder.body[10].mlp_down.x" ] }, { "explicit_count": 0, "missing_count": 5, "status": "warn", "template": "gemma3", "warnings": [ "decoder.footer[2].logits.x", "decoder.body[1].qkv_proj.x", "decoder.body[5].out_proj.x", "decoder.body[9].mlp_gate_up.x", "decoder.body[11].mlp_down.x" ] }, { "explicit_count": 4, "missing_count": 17, "status": "warn", "template": "gemma4", "warnings": [ "decoder.footer[2].logits.x", "decoder.body:sliding_attention_kv[1].q_proj.x", "decoder.body:sliding_attention_kv[2].k_proj.x", "decoder.body:sliding_attention_kv[3].v_proj.x", "decoder.body:sliding_attention_kv[8].out_proj.x", "decoder.body:sliding_attention_kv[12].mlp_gate_up.x", "decoder.body:full_attention_kv[1].q_proj.x", "decoder.body:full_attention_kv[2].k_proj.x" ] }, { "explicit_count": 0, "missing_count": 6, "status": "warn", "template": "gemma4_vision", "warnings": [ "vision_encoder.body[1].q_proj.x", "vision_encoder.body[2].k_proj.x", "vision_encoder.body[3].v_proj.x", "vision_encoder.body[8].out_proj.x", "vision_encoder.body[12].mlp_gate_up.x", "vision_encoder.body[14].mlp_down.x" ] }, { "explicit_count": 5, "missing_count": 0, "status": "pass", "template": "glm4", "warnings": [] }, { "explicit_count": 5, "missing_count": 4, "status": "warn", "template": "nemotron_h", "warnings": [ "decoder.footer[2].logits.x", "decoder.body:mamba[7].mamba_out_proj.x", "decoder.body:attention[5].out_proj.x", "decoder.body:mlp[3].mlp_down.x" ] }, { "explicit_count": 30, "missing_count": 0, "status": "pass", "template": "kimi_vl", "warnings": [] }, { "explicit_count": 0, "missing_count": 5, "status": "warn", "template": "llama", "warnings": [ "decoder.footer[2].logits.x", "decoder.body[1].qkv_proj.x", "decoder.body[4].out_proj.x", "decoder.body[7].mlp_gate_up.x", "decoder.body[9].mlp_down.x" ] } ], "status": "warn", "warnings": 10 }, { "checks_failed": 0, "checks_passed": 4, "details": "Covered by test-v8-template-circuit-audit, GLM4 synthetic BF16/quant dataflow tests, and safetensors-to-BUMP Nemotron checks.", "id": "lowered_ir_dataflow", "label": "Lowered IR Dataflow", "status": "pass", "warnings": 0 }, { "checks_failed": 0, "checks_passed": 1, "details": "Generated-C mamba_in_proj buffer preservation is covered by the template circuit audit test.", "id": "generated_c_preservation", "label": "Generated C Preservation", "status": "pass", "warnings": 0 }, { "checks_failed": 0, "checks_passed": 5, "details": "Mamba2, DeltaNet, sliding attention, KV-cache, and threadpool parity tests exist; long-context Qwen3.5 and Gemma4 vision remain monitored.", "id": "runtime_path_equivalence", "label": "Runtime Path Equivalence", "status": "warn", "warnings": 2 }, { "checks_failed": 0, "checks_passed": 6, "details": "Artifact-backed input storage, compute, reduction, rounding, output storage, shape, thread, and oracle evidence.", "id": "numerical_kernel_capabilities", "label": "Numerical Kernel Capabilities", "rows": [ { "capability": "fp32_fp32_fp32", "compute": "fp32", "head_dim": 8, "input_storage": "fp32", "kernel_family": "vision_mrope", "max_abs": 2.384185791015625e-07, "oracle": "pytorch_formula_and_dtype_cast", "output_storage": "fp32", "reduction": "none", "status": "pass", "threads": "serial_independent_heads" }, { "capability": "fp32_fp32_bf16", "compute": "fp32", "head_dim": 8, "input_storage": "fp32", "kernel_family": "vision_mrope", "max_abs": 0.0, "oracle": "pytorch_formula_and_dtype_cast", "output_storage": "bf16", "reduction": "none", "status": "pass", "threads": "serial_independent_heads" }, { "capability": "fp32_fp32_fp16", "compute": "fp32", "head_dim": 8, "input_storage": "fp32", "kernel_family": "vision_mrope", "max_abs": 0.0, "oracle": "pytorch_formula_and_dtype_cast", "output_storage": "fp16", "reduction": "none", "status": "pass", "threads": "serial_independent_heads" }, { "capability": "fp32_fp32_fp32", "compute": "fp32", "head_dim": 72, "input_storage": "fp32", "kernel_family": "vision_mrope", "max_abs": 2.384185791015625e-07, "oracle": "pytorch_formula_and_dtype_cast", "output_storage": "fp32", "reduction": "none", "status": "pass", "threads": "serial_independent_heads" }, { "capability": "fp32_fp32_bf16", "compute": "fp32", "head_dim": 72, "input_storage": "fp32", "kernel_family": "vision_mrope", "max_abs": 0.0, "oracle": "pytorch_formula_and_dtype_cast", "output_storage": "bf16", "reduction": "none", "status": "pass", "threads": "serial_independent_heads" }, { "capability": "fp32_fp32_fp16", "compute": "fp32", "head_dim": 72, "input_storage": "fp32", "kernel_family": "vision_mrope", "max_abs": 0.0, "oracle": "pytorch_formula_and_dtype_cast", "output_storage": "fp16", "reduction": "none", "status": "pass", "threads": "serial_independent_heads" } ], "status": "pass", "warnings": 0 }, { "checks_failed": 0, "checks_passed": 7, "details": "Promoted family coverage from current v8 bring-up lanes.", "id": "model_contract_coverage", "label": "Model Contract Coverage", "rows": [ { "family": "qwen2", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "fast v8 regression family", "runtime": "pass", "template": "pass" }, { "family": "qwen3", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "QK-norm + GGUF smoke", "runtime": "pass", "template": "pass" }, { "family": "qwen3.5", "generated_c": "pass", "ir": "pass", "model": "partial", "notes": "long-generation numerical parity monitored", "runtime": "partial", "template": "pass" }, { "family": "qwen3-vl", "generated_c": "partial", "ir": "pass", "model": "pass", "notes": "promoted vision smoke; bridge parity still active", "runtime": "partial", "template": "pass" }, { "family": "gemma3", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "split-half RoPE and sliding attention covered", "runtime": "pass", "template": "pass" }, { "family": "gemma4", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "text coherent; vision bridge early", "runtime": "partial", "template": "pass" }, { "family": "glm4", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "partial pairwise RoPE + GGUF smoke", "runtime": "partial", "template": "pass" }, { "family": "nemotron-h", "generated_c": "pass", "ir": "pass", "model": "pass", "notes": "Mamba2 path equivalence monitored", "runtime": "partial", "template": "pass" }, { "family": "kimi-vl", "generated_c": "partial", "ir": "partial", "model": "partial", "notes": "MLA/MoE scalar contracts added; full template lowering/tokenizer/vision bridge in bring-up", "runtime": "partial", "template": "pass" }, { "family": "llama/nanbeige", "generated_c": "pass", "ir": "pass", "model": "partial", "notes": "Nanbeige active bring-up lane", "runtime": "partial", "template": "pass" } ], "status": "warn", "warnings": 7 } ], "status": "warn", "summary": { "explicit_template_edges": 56, "missing_template_edges": 69, "sections_failed": 0, "sections_passed": 3, "sections_total": 6, "sections_warn": 3, "templates_failed": 0, "templates_total": 12, "templates_warn": 10, "warnings": 19 } }, "vision_visualizer": { "schema": "ck.v8.vision_visualizer_harness.v1", "generated_at": "2026-09-21T08:46:53.561391Z", "status": "pass", "summary": { "passed": 2, "failed": 0, "total": 2 }, "checks": [ { "name": "derived_payload", "status": "pass", "detail": "9 multimodal ops" }, { "name": "html_embedding", "status": "pass", "detail": "/tmp/ck-v8-vision-viz-0imccimr/ir_report.html" } ] }, "vision_encoder_accuracy": { "schema_version": 1, "family": "qwen3vl", "mode": "q4", "image": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/vision_encoder_accuracy/fixtures/synthetic_form_1152x896.ppm", "image_geometry": [ 1152, 896 ], "cpu": { "avx512f": false, "avx512_bf16": false }, "layers": [ 0, 8, 16, 24, 26 ], "phases": { "q8_mmproj_llamacpp": { "status": "pass", "report": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/vision_encoder_accuracy/q8_mmproj_llamacpp/summary.json" } }, "started_unix": 1789980472.8011367, "failures": [], "status": "pass", "elapsed_sec": 225.41793131828308 }, "regression_fast": { "status": "pass", "mode": "fast", "generated_at": "2026-09-21T08:02:16", "report_dir": "build/regression-reports/20260921_075455", "summary_path": "build/regression-reports/20260921_075455/summary.json", "families_total": 6, "families_passed": 5, "families_failed": 0, "failure_classes": { "pass": 5 }, "family_rows": [ { "family_id": "gemma", "label": "Gemma 3 270M IT", "status": "PASS", "build_status": "PASS", "smoke_status": "PASS", "coherence_status": "PASS", "contract_status": "PASS", "first_token_status": null, "failure_reason": "", "failure_detail": "" }, { "family_id": "qwen2", "label": "Qwen2 0.5B Instruct", "status": "PASS", "build_status": "PASS", "smoke_status": "PASS", "coherence_status": "PASS", "contract_status": "PASS", "first_token_status": null, "failure_reason": "", "failure_detail": "" }, { "family_id": "qwen3", "label": "Qwen3 0.6B", "status": "PASS", "build_status": "PASS", "smoke_status": "PASS", "coherence_status": "PASS", "contract_status": "PASS", "first_token_status": null, "failure_reason": "", "failure_detail": "" }, { "family_id": "qwen35", "label": "Qwen3.5 0.8B", "status": "PASS", "build_status": "PASS", "smoke_status": "PASS", "coherence_status": "PASS", "contract_status": "PASS", "first_token_status": null, "failure_reason": "", "failure_detail": "" }, { "family_id": "nanbeige", "label": "Nanbeige 4.1 3B", "status": "PASS", "build_status": "PASS", "smoke_status": "PASS", "coherence_status": "FAIL", "contract_status": "PASS", "first_token_status": null, "failure_reason": "coherence_failed:hello:too_short:0<8,too_few_words:0<3,printable_ratio:0.000<0.960,keyword_hits:0<1", "failure_detail": "" }, { "family_id": "gemma4", "label": "Gemma4 E4B IT", "status": "SKIP", "build_status": "SKIP", "smoke_status": "SKIP", "coherence_status": "SKIP", "contract_status": "SKIP", "first_token_status": "SKIP", "failure_detail": "high-memory smoke; skipped unless runner has enough RAM" } ], "note": "Source: build/regression-reports/20260921_075455/summary.json" }, "kernel_map_contracts": { "status": "pass", "generated_at": "2026-09-21T08:14:25.654112Z", "summary_path": "version/v7/.cache/reports/kernel_map_validation_latest.json", "kernel_maps": 112, "passed": 112, "failed": 0, "warnings": 33, "strict": false, "sample_warnings": [ "geglu_forward_bf16.json: Source file not found: src/kernels/bf16_utils.c", "kv_cache_store.json: Source file not found: src/kernels/kv_cache.c", "mega_fused_attention_decode.json: Source file not found: src/kernels/fused/mega_fused_attention.c", "mega_fused_attention_decode_q5_0.json: Fuses unknown kernel 'add_inplace_f32'", "mega_fused_attention_decode_q5_0.json: Fuses unknown kernel 'add_inplace_f32'" ], "note": "Source: version/v7/.cache/reports/kernel_map_validation_latest.json" }, "kernel_allocation": { "status": "warn", "generated_at": "2026-09-21T08:38:29.384332Z", "summary_path": "version/v8/.cache/reports/kernel_allocation_audit_latest.json", "counts": { "allocator_calls": 173, "allocation_calls": 55, "free_calls": 118, "production_allocator_calls": 156, "production_allocation_calls": 51, "frontend_allocator_calls": 3, "oracle_allocator_calls": 14, "test_allocator_calls": 0, "mapped_allocating_providers": 2, "mapped_allocating_without_scratch_contract": 2, "reviewed_allocating_functions": 21, "unreviewed_allocating_functions": 0, "external_library_allocation_risks": 5, "external_library_create_calls": 31 }, "warnings": [ { "code": "production_allocator_debt", "count": 51, "message": "production kernel allocation calls remain" }, { "code": "mapped_allocator_without_scratch", "count": 2, "message": "mapped allocating providers still lack complete scratch ownership" } ], "remaining_providers": [ { "id": "gemm_nt_bf16_pytorch_onednn_brgemm_bf16_storage", "function": "gemm_nt_bf16_pytorch_onednn_brgemm_bf16_storage_impl", "scratch_declared": false, "scratch_bound": false, "selection_status": null }, { "id": "patch_projection_image_bf16_pytorch_onednn_conv3d_storage", "function": "patch_projection_image_bf16_pytorch_onednn_conv3d_storage", "scratch_declared": false, "scratch_bound": false, "selection_status": null } ], "allocation_sites": [ { "path": "src/kernels/attention_kernels.c", "line": 4114, "function": "attention_forward_head_major_gqa_unfused_f16_strict", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4115, "function": "attention_forward_head_major_gqa_unfused_f16_strict", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4326, "function": "attention_forward_head_major_gqa_flash_impl", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4709, "function": "ck_attention_full_bf16_sdpa_amx_range", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4710, "function": "ck_attention_full_bf16_sdpa_amx_range", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4711, "function": "ck_attention_full_bf16_sdpa_amx_range", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4712, "function": "ck_attention_full_bf16_sdpa_amx_range", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4713, "function": "ck_attention_full_bf16_sdpa_amx_range", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4714, "function": "ck_attention_full_bf16_sdpa_amx_range", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4715, "function": "ck_attention_full_bf16_sdpa_amx_range", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4716, "function": "ck_attention_full_bf16_sdpa_amx_range", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 4848, "function": "ck_attention_full_bf16_sdpa_tiled_range", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 5431, "function": "attention_forward_causal_head_major_gqa_flash_strided_f16kv", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 5788, "function": "attention_forward_full_head_major_gqa_ggml_strided", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8034, "function": "attention_forward_causal_head_major_gqa_prefill_append_f16cache_contract", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8353, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8354, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8355, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8357, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8359, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8361, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8363, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8365, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8367, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8369, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8371, "function": "ck_attention_decode_bf16_pytorch_math_gqa", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8512, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8513, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8514, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8515, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8516, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8518, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8520, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8522, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8524, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8526, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 8528, "function": "ck_attention_prefill_bf16_pytorch_math_gqa_full", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_kernels.c", "line": 9292, "function": "attention_forward_causal_head_major_gqa_prefill_append_bf16cache_pytorch_contract", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/attention_oracle_ggml.c", "line": 173, "function": "ck_attention_oracle_dump_tensor", "allocator": "malloc", "classification": "oracle", "provider_ids": [] }, { "path": "src/kernels/attention_oracle_ggml.c", "line": 774, "function": "ck_attention_full_ggml_graph_oracle_multihead", "allocator": "malloc", "classification": "oracle", "provider_ids": [] }, { "path": "src/kernels/attention_oracle_ggml.c", "line": 892, "function": "ck_attention_full_ggml_graph_oracle_multihead", "allocator": "malloc", "classification": "oracle", "provider_ids": [] }, { "path": "src/kernels/audio_kernels.c", "line": 902, "function": "audio_whisper_log_mel_window_wav_pcm16_f32", "allocator": "malloc", "classification": "frontend", "provider_ids": [] }, { "path": "src/kernels/deepseek_kernels.c", "line": 891, "function": "deepseek_mla_attention_f32", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/deepseek_kernels.c", "line": 1058, "function": "deepseek_mla_attention_decode_f32", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/fused/mega_fused_attention_avx.c", "line": 719, "function": "mega_fused_attention_decode", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/gemm_kernels_bf16.c", "line": 1329, "function": "gemm_nt_bf16_amx_bf16_storage", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/gemm_kernels_bf16.c", "line": 1568, "function": "gemm_nt_bf16_pytorch_onednn_brgemm_bf16_storage_impl", "allocator": "malloc", "classification": "production", "provider_ids": [ "gemm_nt_bf16_pytorch_onednn_brgemm_bf16_storage" ] }, { "path": "src/kernels/gemm_kernels_bf16.c", "line": 1569, "function": "gemm_nt_bf16_pytorch_onednn_brgemm_bf16_storage_impl", "allocator": "malloc", "classification": "production", "provider_ids": [ "gemm_nt_bf16_pytorch_onednn_brgemm_bf16_storage" ] }, { "path": "src/kernels/gemm_kernels_bf16.c", "line": 1570, "function": "gemm_nt_bf16_pytorch_onednn_brgemm_bf16_storage_impl", "allocator": "malloc", "classification": "production", "provider_ids": [ "gemm_nt_bf16_pytorch_onednn_brgemm_bf16_storage" ] }, { "path": "src/kernels/gemm_kernels_bf16.c", "line": 1658, "function": "patch_projection_bf16_pytorch_onednn_conv3d_storage", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/gemm_kernels_bf16.c", "line": 1659, "function": "patch_projection_bf16_pytorch_onednn_conv3d_storage", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/gemm_kernels_bf16.c", "line": 1660, "function": "patch_projection_bf16_pytorch_onednn_conv3d_storage", "allocator": "malloc", "classification": "production", "provider_ids": [] }, { "path": "src/kernels/gemm_kernels_bf16.c", "line": 1842, "function": "patch_projection_image_bf16_pytorch_onednn_conv3d_storage", "allocator": "malloc", "classification": "production", "provider_ids": [ "patch_projection_image_bf16_pytorch_onednn_conv3d_storage" ] }, { "path": "src/kernels/gemm_kernels_bf16.c", "line": 1843, "function": "patch_projection_image_bf16_pytorch_onednn_conv3d_storage", "allocator": "malloc", "classification": "production", "provider_ids": [ "patch_projection_image_bf16_pytorch_onednn_conv3d_storage" ] }, { "path": "src/kernels/recurrent_state_kernels.c", "line": 280, "function": "recurrent_conv_state_update_backward", "allocator": "malloc", "classification": "production", "provider_ids": [] } ], "note": "Source: version/v8/.cache/reports/kernel_allocation_audit_latest.json" }, "backprop_regression": { "status": "pass", "mode": "full", "stage_profile": "default", "generated_at": "2026-09-21T08:19:03.489301Z", "summary_path": "version/v7/.cache/reports/backprop_family_matrix/full/v7_backprop_family_matrix_latest.json", "families_total": 5, "families_passed": 5, "families_failed": 0, "family_rows": [ { "family": "qwen2", "status": "PASS", "a_status": "PASS", "b_status": "PASS", "c_status": "PASS", "d_status": "PASS", "e_status": "PASS", "f_status": "PASS", "failed_stage_ids": [], "report_json": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v7/.cache/reports/backprop_family_matrix/full/qwen2/training_parity_regimen_latest.json" }, { "family": "qwen3", "status": "PASS", "a_status": "PASS", "b_status": "PASS", "c_status": "PASS", "d_status": "PASS", "e_status": "PASS", "f_status": "PASS", "failed_stage_ids": [], "report_json": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v7/.cache/reports/backprop_family_matrix/full/qwen3/training_parity_regimen_latest.json" }, { "family": "gemma", "status": "PASS", "a_status": "PASS", "b_status": "PASS", "c_status": "PASS", "d_status": "PASS", "e_status": "PASS", "f_status": "PASS", "failed_stage_ids": [], "report_json": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v7/.cache/reports/backprop_family_matrix/full/gemma/training_parity_regimen_latest.json" }, { "family": "nanbeige", "status": "PASS", "a_status": "PASS", "b_status": "PASS", "c_status": "PASS", "d_status": "PASS", "e_status": "PASS", "f_status": "PASS", "failed_stage_ids": [], "report_json": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v7/.cache/reports/backprop_family_matrix/full/nanbeige/training_parity_regimen_latest.json" }, { "family": "qwen35", "status": "PASS", "a_status": "PASS", "b_status": "PASS", "c_status": "PASS", "d_status": "PASS", "e_status": "PASS", "f_status": "PASS", "failed_stage_ids": [], "report_json": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v7/.cache/reports/backprop_family_matrix/full/qwen35/training_parity_regimen_latest.json" } ], "failed_families": [], "note": "Source: version/v7/.cache/reports/backprop_family_matrix/full/v7_backprop_family_matrix_latest.json" }, "training_certification": { "status": "pass", "passed": true, "generated_at": "2026-09-21T08:19:03.954784Z", "summary_path": "version/v8/.cache/reports/training_certification_latest.json", "execution": { "actual_runtime_threads": 2, "completed_at": "2026-09-21T08:19:14.595562Z", "cpu_affinity": [ 0, 1, 2, 3 ], "cpu_affinity_count": 4, "git_commit": "19694d9d9ba7cef51bce0e5ee7229041802bcb9e", "github": { "run_attempt": "1", "run_id": "35574289723", "sha": "19694d9d9ba7cef51bce0e5ee7229041802bcb9e" }, "hostname": "runnervmlun5p", "pid": 20388, "run_id": "be0f5933-d516-4254-bc6b-a73ac0b6f293", "started_at": "2026-09-21T08:19:03.954839Z", "thread_environment": { "CK_NUM_THREADS": null, "OMP_NUM_THREADS": null } }, "configuration": { "d_model": 64, "dtype": "fp32", "heads": 4, "hidden": 128, "kv_heads": 4, "layers": 2, "model": "fp32_dense_2layer", "seed": 42, "sequence_length": 8, "vocab": 256 }, "checks": { "generated_execution": { "backward_calls": 109, "forward_calls": 31, "logits_slot": "act.Sfooter.logits.0.y", "passed": true }, "gradients": { "compared": 35, "failed": [], "passed": true, "tensors": [ { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 2.3739412426948547e-06, "mean_abs_diff": 6.377358090503549e-07, "name": "final_ln_weight", "numel": 64, "passed": true, "reference_max_abs": 0.022124195471405983, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 2.3739412426948547e-06, "worst_actual": -0.0101793697103858, "worst_allowed_error": 0.0003305452410131693, "worst_error_ratio": 0.007181894965469837, "worst_expected": -0.010181743651628494, "worst_index": 51 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 2.1506566554307938e-06, "mean_abs_diff": 5.336872277439397e-07, "name": "layer.0.b1", "numel": 256, "passed": true, "reference_max_abs": 0.012535997666418552, "relative_tolerance": 0.003, "shape": [ 256 ], "violations": 0, "worst_abs_diff": 2.1506566554307938e-06, "worst_actual": 0.0007850149995647371, "worst_allowed_error": 0.00030234860605560243, "worst_error_ratio": 0.007113168947398663, "worst_expected": 0.0007828643429093063, "worst_index": 153 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 0.00010959059000015259, "mean_abs_diff": 2.961735299322754e-05, "name": "layer.0.b2", "numel": 64, "passed": true, "reference_max_abs": 0.6261330842971802, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 8.845329284667969e-05, "worst_actual": 0.01327560842037201, "worst_allowed_error": 0.00033956149127334356, "worst_error_ratio": 0.2604927122592926, "worst_expected": 0.01318715512752533, "worst_index": 3 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 1.2017786502838135e-05, "mean_abs_diff": 3.005204462169786e-06, "name": "layer.0.bk", "numel": 64, "passed": true, "reference_max_abs": 0.04105398803949356, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 1.1431984603404999e-05, "worst_actual": -0.013749107718467712, "worst_allowed_error": 0.00034128164406865835, "worst_error_ratio": 0.03349721431732178, "worst_expected": -0.013760539703071117, "worst_index": 3 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 0.00010744482278823853, "mean_abs_diff": 2.932947245426476e-05, "name": "layer.0.bo", "numel": 64, "passed": true, "reference_max_abs": 0.6823280453681946, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 0.0001010596752166748, "worst_actual": 0.02111433446407318, "worst_allowed_error": 0.0003630398423410952, "worst_error_ratio": 0.27837076783180237, "worst_expected": 0.021013274788856506, "worst_index": 3 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 2.291100099682808e-05, "mean_abs_diff": 6.97708128427621e-06, "name": "layer.0.bq", "numel": 64, "passed": true, "reference_max_abs": 0.051279157400131226, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 2.291100099682808e-05, "worst_actual": -0.006140823010355234, "worst_allowed_error": 0.0003184912202414125, "worst_error_ratio": 0.07193604856729507, "worst_expected": -0.006163734011352062, "worst_index": 2 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 2.2598542273044586e-05, "mean_abs_diff": 5.67990900890436e-06, "name": "layer.0.bv", "numel": 64, "passed": true, "reference_max_abs": 0.09422754496335983, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 2.2598542273044586e-05, "worst_actual": -0.0008644340559840202, "worst_allowed_error": 0.0003025255282409489, "worst_error_ratio": 0.07469961792230606, "worst_expected": -0.0008418355137109756, "worst_index": 16 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 4.130415618419647e-06, "mean_abs_diff": 2.1822643248015083e-06, "name": "layer.0.k_norm", "numel": 16, "passed": true, "reference_max_abs": 0.009572567418217659, "relative_tolerance": 0.003, "shape": [ 16 ], "violations": 0, "worst_abs_diff": 4.130415618419647e-06, "worst_actual": 0.00956843700259924, "worst_allowed_error": 0.0003287177241872996, "worst_error_ratio": 0.01256523560732603, "worst_expected": 0.009572567418217659, "worst_index": 9 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 5.964189767837524e-06, "mean_abs_diff": 1.6055945479820366e-06, "name": "layer.0.ln1_gamma", "numel": 64, "passed": true, "reference_max_abs": 0.017056645825505257, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 5.647540092468262e-06, "worst_actual": 0.007749788463115692, "worst_allowed_error": 0.0003232663148082793, "worst_error_ratio": 0.017470240592956543, "worst_expected": 0.00775543600320816, "worst_index": 34 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 1.2547243386507034e-06, "mean_abs_diff": 2.027745154009608e-07, "name": "layer.0.ln2_gamma", "numel": 64, "passed": true, "reference_max_abs": 0.0035797960590571165, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 1.2547243386507034e-06, "worst_actual": -0.001638480694964528, "worst_allowed_error": 0.0003049116930924356, "worst_error_ratio": 0.004115041811019182, "worst_expected": -0.0016372259706258774, "worst_index": 11 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 5.395151674747467e-06, "mean_abs_diff": 2.149586180166807e-06, "name": "layer.0.q_norm", "numel": 16, "passed": true, "reference_max_abs": 0.011954184621572495, "relative_tolerance": 0.003, "shape": [ 16 ], "violations": 0, "worst_abs_diff": 5.395151674747467e-06, "worst_actual": 0.011948789469897747, "worst_allowed_error": 0.00033586256904527545, "worst_error_ratio": 0.016063569113612175, "worst_expected": 0.011954184621572495, "worst_index": 9 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 5.4622069001197815e-06, "mean_abs_diff": 6.045275426913577e-07, "name": "layer.0.w1", "numel": 16384, "passed": true, "reference_max_abs": 0.02320639044046402, "relative_tolerance": 0.003, "shape": [ 256, 64 ], "violations": 0, "worst_abs_diff": 4.997171345166862e-06, "worst_actual": 7.452173304045573e-05, "worst_allowed_error": 0.0003002086014021188, "worst_error_ratio": 0.01664566434919834, "worst_expected": 6.952456169528887e-05, "worst_index": 5990 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 9.933486580848694e-06, "mean_abs_diff": 6.04732804276864e-07, "name": "layer.0.w2", "numel": 8192, "passed": true, "reference_max_abs": 0.03104834072291851, "relative_tolerance": 0.003, "shape": [ 64, 128 ], "violations": 0, "worst_abs_diff": 9.933486580848694e-06, "worst_actual": -0.02195853181183338, "worst_allowed_error": 0.0003659054054878652, "worst_error_ratio": 0.0271476898342371, "worst_expected": -0.02196846529841423, "worst_index": 5300 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 4.0978193283081055e-05, "mean_abs_diff": 5.733709258493036e-06, "name": "layer.0.wk", "numel": 4096, "passed": true, "reference_max_abs": 0.0921480804681778, "relative_tolerance": 0.003, "shape": [ 64, 64 ], "violations": 0, "worst_abs_diff": 3.8623809814453125e-05, "worst_actual": 0.042841874063014984, "worst_allowed_error": 0.00042864150600507855, "worst_error_ratio": 0.09010748565196991, "worst_expected": 0.04288049787282944, "worst_index": 1398 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 3.7789344787597656e-05, "mean_abs_diff": 4.973216164216865e-06, "name": "layer.0.wo", "numel": 4096, "passed": true, "reference_max_abs": 0.2272457778453827, "relative_tolerance": 0.003, "shape": [ 64, 64 ], "violations": 0, "worst_abs_diff": 2.3276079446077347e-05, "worst_actual": -0.001974430400878191, "worst_allowed_error": 0.0003059931332245469, "worst_error_ratio": 0.07606732845306396, "worst_expected": -0.0019977064803242683, "worst_index": 2679 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 6.3285231590271e-05, "mean_abs_diff": 6.459388714574743e-06, "name": "layer.0.wq", "numel": 4096, "passed": true, "reference_max_abs": 0.14635930955410004, "relative_tolerance": 0.003, "shape": [ 64, 64 ], "violations": 0, "worst_abs_diff": 6.160640623420477e-05, "worst_actual": -0.0019727805629372597, "worst_allowed_error": 0.00030573352705687284, "worst_error_ratio": 0.20150360465049744, "worst_expected": -0.001911174156703055, "worst_index": 162 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 4.703551530838013e-05, "mean_abs_diff": 5.187218903301982e-06, "name": "layer.0.wv", "numel": 4096, "passed": true, "reference_max_abs": 0.148093119263649, "relative_tolerance": 0.003, "shape": [ 64, 64 ], "violations": 0, "worst_abs_diff": 4.703551530838013e-05, "worst_actual": 0.020636485889554024, "worst_allowed_error": 0.0003620505740400404, "worst_error_ratio": 0.12991420924663544, "worst_expected": 0.020683521404862404, "worst_index": 2650 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 1.8621794879436493e-06, "mean_abs_diff": 3.1820599133425276e-07, "name": "layer.1.b1", "numel": 256, "passed": true, "reference_max_abs": 0.01062451209872961, "relative_tolerance": 0.003, "shape": [ 256 ], "violations": 0, "worst_abs_diff": 1.8621794879436493e-06, "worst_actual": 0.006018395535647869, "worst_allowed_error": 0.00031806077458895743, "worst_error_ratio": 0.00585479149594903, "worst_expected": 0.006020257715135813, "worst_index": 222 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 2.1725893020629883e-05, "mean_abs_diff": 6.985792424529791e-06, "name": "layer.1.b2", "numel": 64, "passed": true, "reference_max_abs": 0.4930659830570221, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 1.3407319784164429e-05, "worst_actual": 0.03530207648873329, "worst_allowed_error": 0.00040594645543023944, "worst_error_ratio": 0.03302730992436409, "worst_expected": 0.035315483808517456, "worst_index": 54 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 4.798173904418945e-06, "mean_abs_diff": 1.5410678315674886e-06, "name": "layer.1.bk", "numel": 64, "passed": true, "reference_max_abs": 0.021448882296681404, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 4.798173904418945e-06, "worst_actual": -0.009464508853852749, "worst_allowed_error": 0.00032840794301591814, "worst_error_ratio": 0.014610407873988152, "worst_expected": -0.009469307027757168, "worst_index": 49 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 3.0487775802612305e-05, "mean_abs_diff": 8.05771560408175e-06, "name": "layer.1.bo", "numel": 64, "passed": true, "reference_max_abs": 0.46859169006347656, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 2.421438694000244e-05, "worst_actual": -0.08486391603946686, "worst_allowed_error": 0.0005546644097194076, "worst_error_ratio": 0.04365592449903488, "worst_expected": -0.08488813042640686, "worst_index": 8 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 9.469687938690186e-06, "mean_abs_diff": 3.349059170432156e-06, "name": "layer.1.bq", "numel": 64, "passed": true, "reference_max_abs": 0.03248206526041031, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 9.3802809715271e-06, "worst_actual": -0.01654980331659317, "worst_allowed_error": 0.0003496775752864778, "worst_error_ratio": 0.026825515553355217, "worst_expected": -0.016559183597564697, "worst_index": 51 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 3.9711594581604e-06, "mean_abs_diff": 1.5672285371692851e-06, "name": "layer.1.bv", "numel": 64, "passed": true, "reference_max_abs": 0.07385999709367752, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 3.5243574529886246e-06, "worst_actual": 0.0011103337164968252, "worst_allowed_error": 0.00030334159964695573, "worst_error_ratio": 0.011618443764746189, "worst_expected": 0.0011138580739498138, "worst_index": 7 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 1.5273690223693848e-06, "mean_abs_diff": 7.553339855803642e-07, "name": "layer.1.k_norm", "numel": 16, "passed": true, "reference_max_abs": 0.0054721105843782425, "relative_tolerance": 0.003, "shape": [ 16 ], "violations": 0, "worst_abs_diff": 1.5273690223693848e-06, "worst_actual": 0.005470583215355873, "worst_allowed_error": 0.00031641634996049106, "worst_error_ratio": 0.0048270863480865955, "worst_expected": 0.0054721105843782425, "worst_index": 15 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 4.3264590203762054e-06, "mean_abs_diff": 8.680573273522896e-07, "name": "layer.1.ln1_gamma", "numel": 64, "passed": true, "reference_max_abs": 0.016144661232829094, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 4.3264590203762054e-06, "worst_actual": -0.0029326719231903553, "worst_allowed_error": 0.0003087850636802614, "worst_error_ratio": 0.014011231251060963, "worst_expected": -0.002928345464169979, "worst_index": 17 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 4.423345671966672e-07, "mean_abs_diff": 1.5018045473880193e-07, "name": "layer.1.ln2_gamma", "numel": 64, "passed": true, "reference_max_abs": 0.002671240596100688, "relative_tolerance": 0.003, "shape": [ 64 ], "violations": 0, "worst_abs_diff": 4.423345671966672e-07, "worst_actual": -6.17823243374005e-05, "worst_allowed_error": 0.0003001840377692133, "worst_error_ratio": 0.001473544631153345, "worst_expected": -6.133998977020383e-05, "worst_index": 8 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 1.5264376997947693e-06, "mean_abs_diff": 8.118163350445684e-07, "name": "layer.1.q_norm", "numel": 16, "passed": true, "reference_max_abs": 0.006321600638329983, "relative_tolerance": 0.003, "shape": [ 16 ], "violations": 0, "worst_abs_diff": 1.5264376997947693e-06, "worst_actual": 0.005470601376146078, "worst_allowed_error": 0.000316416408168152, "worst_error_ratio": 0.004824141971766949, "worst_expected": 0.005472127813845873, "worst_index": 15 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 4.898756742477417e-06, "mean_abs_diff": 4.1714713461260544e-07, "name": "layer.1.w1", "numel": 16384, "passed": true, "reference_max_abs": 0.02459542267024517, "relative_tolerance": 0.003, "shape": [ 256, 64 ], "violations": 0, "worst_abs_diff": 4.520639777183533e-06, "worst_actual": -0.013268304988741875, "worst_allowed_error": 0.0003398184780962765, "worst_error_ratio": 0.013303101994097233, "worst_expected": -0.013272825628519058, "worst_index": 14267 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 4.28222119808197e-06, "mean_abs_diff": 4.430377202879754e-07, "name": "layer.1.w2", "numel": 8192, "passed": true, "reference_max_abs": 0.027457494288682938, "relative_tolerance": 0.003, "shape": [ 64, 128 ], "violations": 0, "worst_abs_diff": 4.176050424575806e-06, "worst_actual": -0.002041744301095605, "worst_allowed_error": 0.00030611272086389363, "worst_error_ratio": 0.013642198406159878, "worst_expected": -0.002037568250671029, "worst_index": 7312 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 2.483278512954712e-05, "mean_abs_diff": 3.100064986938378e-06, "name": "layer.1.wk", "numel": 4096, "passed": true, "reference_max_abs": 0.05050027742981911, "relative_tolerance": 0.003, "shape": [ 64, 64 ], "violations": 0, "worst_abs_diff": 2.2413209080696106e-05, "worst_actual": 0.02206234820187092, "worst_allowed_error": 0.00036625430220738053, "worst_error_ratio": 0.06119575724005699, "worst_expected": 0.022084761410951614, "worst_index": 3650 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 2.5019049644470215e-05, "mean_abs_diff": 3.008502062584739e-06, "name": "layer.1.wo", "numel": 4096, "passed": true, "reference_max_abs": 0.1375257819890976, "relative_tolerance": 0.003, "shape": [ 64, 64 ], "violations": 0, "worst_abs_diff": 1.9352417439222336e-05, "worst_actual": -0.005341240670531988, "worst_allowed_error": 0.00031596567714586854, "worst_error_ratio": 0.06124848127365112, "worst_expected": -0.005321888253092766, "worst_index": 511 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 2.5823712348937988e-05, "mean_abs_diff": 3.071157152589876e-06, "name": "layer.1.wq", "numel": 4096, "passed": true, "reference_max_abs": 0.06176643818616867, "relative_tolerance": 0.003, "shape": [ 64, 64 ], "violations": 0, "worst_abs_diff": 2.3271888494491577e-05, "worst_actual": 0.017791643738746643, "worst_allowed_error": 0.0003534447751007974, "worst_error_ratio": 0.06584306806325912, "worst_expected": 0.017814915627241135, "worst_index": 3446 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 2.5909394025802612e-05, "mean_abs_diff": 3.070633738389006e-06, "name": "layer.1.wv", "numel": 4096, "passed": true, "reference_max_abs": 0.12831151485443115, "relative_tolerance": 0.003, "shape": [ 64, 64 ], "violations": 0, "worst_abs_diff": 2.5909394025802612e-05, "worst_actual": -0.010253389365971088, "worst_allowed_error": 0.0003306824655737728, "worst_error_ratio": 0.07835128158330917, "worst_expected": -0.010227479971945286, "worst_index": 1272 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 5.729496479034424e-05, "mean_abs_diff": 5.770513666902843e-07, "name": "output.weight", "numel": 16384, "passed": true, "reference_max_abs": 0.4397601783275604, "relative_tolerance": 0.003, "shape": [ 256, 64 ], "violations": 0, "worst_abs_diff": 5.1987357437610626e-05, "worst_actual": 0.011786640621721745, "worst_allowed_error": 0.00033551588421687484, "worst_error_ratio": 0.15494753420352936, "worst_expected": 0.011838627979159355, "worst_index": 13528 }, { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 0.00015397369861602783, "mean_abs_diff": 1.067175389835029e-06, "name": "token_emb", "numel": 16384, "passed": true, "reference_max_abs": 0.47191083431243896, "relative_tolerance": 0.003, "shape": [ 256, 64 ], "violations": 0, "worst_abs_diff": 0.00015397369861602783, "worst_actual": 0.04856546223163605, "worst_allowed_error": 0.0004461583448573947, "worst_error_ratio": 0.3451099693775177, "worst_expected": 0.048719435930252075, "worst_index": 9608 } ] }, "loaded_engine_dependency": { "command": [ "/opt/hostedtoolcache/Python/3.11.16/x64/bin/python", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/scripts/run_training_certification_v8.py", "--probe-loaded-dependencies", "--probe-library", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "--probe-json-out", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/loaded_dependency_probe.json", "--probe-runtime-contract", "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "--probe-build-identity", "55db006254289a732c3283e0da1271e4ee5812efbcc49dacf0ed6bbe2a1a2fe8", "--probe-engine-sha256", "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525944, "inode": 9176195, "mtime_ns": 1789977703702379527, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" } ], "loaded_sha256": [ "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ], "observed_in_process_maps": true, "passed": true, "process_map_entries": [ "7fa83a07d000-7fa83a094000 r--p 00000000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7fa83a094000-7fa83a1a4000 r-xp 00017000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7fa83a1a4000-7fa83a1d6000 r--p 00127000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7fa83a1d6000-7fa83a1da000 r--p 00159000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7fa83a1da000-7fa83a1db000 rw-p 0015d000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ] }, "passed": true, "returncode": 0, "stdout": "", "training_build_identity": { "expected": "55db006254289a732c3283e0da1271e4ee5812efbcc49dacf0ed6bbe2a1a2fe8", "loaded": "55db006254289a732c3283e0da1271e4ee5812efbcc49dacf0ed6bbe2a1a2fe8", "passed": true }, "training_contract": { "expected": "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "loaded": "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "passed": true }, "training_library": { "bytes": 200120, "inode": 9437727, "mtime_ns": 1789978750625853437, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "process_map_entries": [ "7fa845ed4000-7fa845eda000 r--p 00000000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7fa845eda000-7fa845ef7000 r-xp 00006000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7fa845ef7000-7fa845efb000 r--p 00023000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7fa845efb000-7fa845eff000 r--p 00027000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7fa845eff000-7fa845f00000 rw-p 0002b000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so" ], "sha256": "e6dae1533c34cf51a06e04a571f1b2d56fd81cc773dc56f343e4b2e37e7786d2" } }, "loaded_library_build_identity": { "expected": "55db006254289a732c3283e0da1271e4ee5812efbcc49dacf0ed6bbe2a1a2fe8", "loaded": "55db006254289a732c3283e0da1271e4ee5812efbcc49dacf0ed6bbe2a1a2fe8", "passed": true }, "loaded_library_contract": { "expected": "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "loaded": "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "passed": true }, "loaded_library_observed": { "passed": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "process_map_entries": [ "7fe8db95c000-7fe8db962000 r--p 00000000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7fe8db962000-7fe8db97f000 r-xp 00006000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7fe8db97f000-7fe8db983000 r--p 00023000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7fe8db983000-7fe8db987000 r--p 00027000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7fe8db987000-7fe8db988000 rw-p 0002b000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so" ] }, "logits": { "abs_tolerance": 0.0002, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 8.52346420288086e-05, "mean_abs_diff": 1.606245496077463e-05, "numel": 2048, "passed": true, "reference_max_abs": 0.5186089277267456, "relative_tolerance": 0.0002, "violations": 0, "worst_abs_diff": 8.52346420288086e-05, "worst_actual": -0.13557133078575134, "worst_allowed_error": 0.00022709721815772355, "worst_error_ratio": 0.37532225251197815, "worst_expected": -0.13548609614372253, "worst_index": 1811 }, "loss": { "abs_diff": 2.86102294921875e-06, "cke": 5.534800052642822, "passed": true, "pytorch": 5.534797191619873, "tolerance": 2e-05 }, "parameter_inventory": { "agreement": { "expected": [ "token_emb", "final_ln_weight", "output.weight", "layer.0.ln1_gamma", "layer.0.ln2_gamma", "layer.0.wq", "layer.0.wk", "layer.0.wv", "layer.0.wo", "layer.0.bq", "layer.0.bk", "layer.0.bv", "layer.0.bo", "layer.0.q_norm", "layer.0.k_norm", "layer.0.w1", "layer.0.w2", "layer.0.b1", "layer.0.b2", "layer.1.ln1_gamma", "layer.1.ln2_gamma", "layer.1.wq", "layer.1.wk", "layer.1.wv", "layer.1.wo", "layer.1.bq", "layer.1.bk", "layer.1.bv", "layer.1.bo", "layer.1.q_norm", "layer.1.k_norm", "layer.1.w1", "layer.1.w2", "layer.1.b1", "layer.1.b2" ], "exported_gradients": [ "final_ln_weight", "layer.0.b1", "layer.0.b2", "layer.0.bk", "layer.0.bo", "layer.0.bq", "layer.0.bv", "layer.0.k_norm", "layer.0.ln1_gamma", "layer.0.ln2_gamma", "layer.0.q_norm", "layer.0.w1", "layer.0.w2", "layer.0.wk", "layer.0.wo", "layer.0.wq", "layer.0.wv", "layer.1.b1", "layer.1.b2", "layer.1.bk", "layer.1.bo", "layer.1.bq", "layer.1.bv", "layer.1.k_norm", "layer.1.ln1_gamma", "layer.1.ln2_gamma", "layer.1.q_norm", "layer.1.w1", "layer.1.w2", "layer.1.wk", "layer.1.wo", "layer.1.wq", "layer.1.wv", "output.weight", "token_emb" ], "generated_inventory": [ "final_ln_weight", "layer.0.b1", "layer.0.b2", "layer.0.bk", "layer.0.bo", "layer.0.bq", "layer.0.bv", "layer.0.k_norm", "layer.0.ln1_gamma", "layer.0.ln2_gamma", "layer.0.q_norm", "layer.0.w1", "layer.0.w2", "layer.0.wk", "layer.0.wo", "layer.0.wq", "layer.0.wv", "layer.1.b1", "layer.1.b2", "layer.1.bk", "layer.1.bo", "layer.1.bq", "layer.1.bv", "layer.1.k_norm", "layer.1.ln1_gamma", "layer.1.ln2_gamma", "layer.1.q_norm", "layer.1.w1", "layer.1.w2", "layer.1.wk", "layer.1.wo", "layer.1.wq", "layer.1.wv", "output.weight", "token_emb" ], "pytorch_gradients": [ "final_ln_weight", "layer.0.b1", "layer.0.b2", "layer.0.bk", "layer.0.bo", "layer.0.bq", "layer.0.bv", "layer.0.k_norm", "layer.0.ln1_gamma", "layer.0.ln2_gamma", "layer.0.q_norm", "layer.0.w1", "layer.0.w2", "layer.0.wk", "layer.0.wo", "layer.0.wq", "layer.0.wv", "layer.1.b1", "layer.1.b2", "layer.1.bk", "layer.1.bo", "layer.1.bq", "layer.1.bv", "layer.1.k_norm", "layer.1.ln1_gamma", "layer.1.ln2_gamma", "layer.1.q_norm", "layer.1.w1", "layer.1.w2", "layer.1.wk", "layer.1.wo", "layer.1.wq", "layer.1.wv", "output.weight", "token_emb" ] }, "count": 35, "expected_source": { "excluded_fixture": [ "tiny.embedding.weight", "tiny.fc1.bias", "tiny.fc1.weight", "tiny.fc2.bias", "tiny.fc2.weight", "tiny.rms_gamma" ], "frozen_by_dtype": [], "policy": "all_primary_fp32_manifest_parameters_excluding_tiny_parity_fixture_and_aliases", "source": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/weights_manifest.json", "tied": [] }, "parameters": [ { "manifest_name": "final_ln_weight", "name": "final_ln_weight", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.0.b1", "name": "layer.0.b1", "numel": 256, "shape": [ 256 ] }, { "manifest_name": "layer.0.b2", "name": "layer.0.b2", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.0.bk", "name": "layer.0.bk", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.0.bo", "name": "layer.0.bo", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.0.bq", "name": "layer.0.bq", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.0.bv", "name": "layer.0.bv", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.0.k_norm", "name": "layer.0.k_norm", "numel": 16, "shape": [ 16 ] }, { "manifest_name": "layer.0.ln1_gamma", "name": "layer.0.ln1_gamma", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.0.ln2_gamma", "name": "layer.0.ln2_gamma", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.0.q_norm", "name": "layer.0.q_norm", "numel": 16, "shape": [ 16 ] }, { "manifest_name": "layer.0.w1", "name": "layer.0.w1", "numel": 16384, "shape": [ 256, 64 ] }, { "manifest_name": "layer.0.w2", "name": "layer.0.w2", "numel": 8192, "shape": [ 64, 128 ] }, { "manifest_name": "layer.0.wk", "name": "layer.0.wk", "numel": 4096, "shape": [ 64, 64 ] }, { "manifest_name": "layer.0.wo", "name": "layer.0.wo", "numel": 4096, "shape": [ 64, 64 ] }, { "manifest_name": "layer.0.wq", "name": "layer.0.wq", "numel": 4096, "shape": [ 64, 64 ] }, { "manifest_name": "layer.0.wv", "name": "layer.0.wv", "numel": 4096, "shape": [ 64, 64 ] }, { "manifest_name": "layer.1.b1", "name": "layer.1.b1", "numel": 256, "shape": [ 256 ] }, { "manifest_name": "layer.1.b2", "name": "layer.1.b2", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.1.bk", "name": "layer.1.bk", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.1.bo", "name": "layer.1.bo", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.1.bq", "name": "layer.1.bq", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.1.bv", "name": "layer.1.bv", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.1.k_norm", "name": "layer.1.k_norm", "numel": 16, "shape": [ 16 ] }, { "manifest_name": "layer.1.ln1_gamma", "name": "layer.1.ln1_gamma", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.1.ln2_gamma", "name": "layer.1.ln2_gamma", "numel": 64, "shape": [ 64 ] }, { "manifest_name": "layer.1.q_norm", "name": "layer.1.q_norm", "numel": 16, "shape": [ 16 ] }, { "manifest_name": "layer.1.w1", "name": "layer.1.w1", "numel": 16384, "shape": [ 256, 64 ] }, { "manifest_name": "layer.1.w2", "name": "layer.1.w2", "numel": 8192, "shape": [ 64, 128 ] }, { "manifest_name": "layer.1.wk", "name": "layer.1.wk", "numel": 4096, "shape": [ 64, 64 ] }, { "manifest_name": "layer.1.wo", "name": "layer.1.wo", "numel": 4096, "shape": [ 64, 64 ] }, { "manifest_name": "layer.1.wq", "name": "layer.1.wq", "numel": 4096, "shape": [ 64, 64 ] }, { "manifest_name": "layer.1.wv", "name": "layer.1.wv", "numel": 4096, "shape": [ 64, 64 ] }, { "manifest_name": "output.weight", "name": "output.weight", "numel": 16384, "shape": [ 256, 64 ] }, { "manifest_name": "token_emb", "name": "token_emb", "numel": 16384, "shape": [ 256, 64 ] } ], "passed": true, "total_gradient_floats": 116224 }, "serialized_weights": { "abs_tolerance": 0.0, "contract": "elementwise_abs_plus_relative", "first_violating_index": null, "max_abs_diff": 0.0, "mean_abs_diff": 0.0, "numel": 116224, "passed": true, "reference_max_abs": 1.0, "relative_tolerance": 0.0, "violations": 0, "worst_abs_diff": 0.0, "worst_actual": 1.0, "worst_allowed_error": 0.0, "worst_error_ratio": 0.0, "worst_expected": 1.0, "worst_index": 0 } }, "negative_controls": { "alternate_engine": { "detector_result": { "command": [ "/opt/hostedtoolcache/Python/3.11.16/x64/bin/python", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/scripts/run_training_certification_v8.py", "--probe-loaded-dependencies", "--probe-library", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "--probe-json-out", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/alternate_engine_dependency_probe.json", "--probe-runtime-contract", "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "--probe-build-identity", "55db006254289a732c3283e0da1271e4ee5812efbcc49dacf0ed6bbe2a1a2fe8", "--probe-engine-sha256", "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525983, "inode": 9437734, "mtime_ns": 1789978754412733733, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/alternate_engine_control/libckernel_engine.so", "sha256": "18b66f9c470745f80a0d1311bb554846cbef3031f2648108d7602cdb2a038d46" } ], "loaded_sha256": [ "18b66f9c470745f80a0d1311bb554846cbef3031f2648108d7602cdb2a038d46" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/alternate_engine_control/libckernel_engine.so" ], "observed_in_process_maps": true, "passed": false, "process_map_entries": [ "7f5b64b7d000-7f5b64b94000 r--p 00000000 08:01 9437734 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/alternate_engine_control/libckernel_engine.so", "7f5b64b94000-7f5b64ca4000 r-xp 00017000 08:01 9437734 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/alternate_engine_control/libckernel_engine.so", "7f5b64ca4000-7f5b64cd6000 r--p 00127000 08:01 9437734 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/alternate_engine_control/libckernel_engine.so", "7f5b64cd6000-7f5b64cda000 r--p 00159000 08:01 9437734 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/alternate_engine_control/libckernel_engine.so", "7f5b64cda000-7f5b64cdb000 rw-p 0015d000 08:01 9437734 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/alternate_engine_control/libckernel_engine.so" ] }, "passed": false, "returncode": 1, "stdout": "", "training_build_identity": { "expected": "55db006254289a732c3283e0da1271e4ee5812efbcc49dacf0ed6bbe2a1a2fe8", "loaded": "55db006254289a732c3283e0da1271e4ee5812efbcc49dacf0ed6bbe2a1a2fe8", "passed": true }, "training_contract": { "expected": "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "loaded": "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "passed": true }, "training_library": { "bytes": 200120, "inode": 9437727, "mtime_ns": 1789978750625853437, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "process_map_entries": [ "7f5b70ad4000-7f5b70ada000 r--p 00000000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7f5b70ada000-7f5b70af7000 r-xp 00006000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7f5b70af7000-7f5b70afb000 r--p 00023000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7f5b70afb000-7f5b70aff000 r--p 00027000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so", "7f5b70aff000-7f5b70b00000 rw-p 0002b000 08:01 9437727 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain.so" ], "sha256": "e6dae1533c34cf51a06e04a571f1b2d56fd81cc773dc56f343e4b2e37e7786d2" } }, "expected_engine": { "bytes": 1525944, "inode": 9176195, "mtime_ns": 1789977703702379527, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" }, "injected_engine": { "bytes": 1525983, "inode": 9437734, "mtime_ns": 1789978754412733733, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/alternate_engine_control/libckernel_engine.so", "sha256": "18b66f9c470745f80a0d1311bb554846cbef3031f2648108d7602cdb2a038d46" }, "passed": true }, "corrupted_gradient": { "detector_result": { "abs_tolerance": 0.0003, "contract": "elementwise_abs_plus_relative", "first_violating_index": 0, "max_abs_diff": 0.9999997615814209, "mean_abs_diff": 0.01562562771141529, "numel": 64, "passed": false, "reference_max_abs": 0.022124195471405983, "relative_tolerance": 0.003, "violations": 1, "worst_abs_diff": 0.9999997615814209, "worst_actual": 0.9971669316291809, "worst_allowed_error": 0.0003084984782617539, "worst_error_ratio": 3241.50634765625, "worst_expected": -0.002832818077877164, "worst_index": 0 }, "injected_parameter": "final_ln_weight", "passed": true }, "stale_library": { "build_identity_detector": { "expected": "55db006254289a732c3283e0da1271e4ee5812efbcc49dacf0ed6bbe2a1a2fe8", "loaded": "66221332d82d46e3fc675baa113ddf06af088c436ea42d20da39f31fd9e8831b", "passed": false }, "passed": true, "same_runtime_contract": { "expected": "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "loaded": "4fab2723cdd1ee358faa0561dcebffbee5eade5668bb88bd4757a442a194717f", "passed": true }, "substituted_build_identity": { "build_identity_sha256": "66221332d82d46e3fc675baa113ddf06af088c436ea42d20da39f31fd9e8831b", "compile_command": [ "gcc", "-shared", "-fPIC", "-O3", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/generated_train_runtime_stale_control.c", "-o", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain_stale_control.so", "-I", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/include", "-I", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine", "-L", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build", "-lckernel_engine", "-lm", "-Wl,-rpath,/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build", "-DCK_NUM_TOKENS=8", "-DCK_TRAIN_BUILD_IDENTITY_SHA256=\"66221332d82d46e3fc675baa113ddf06af088c436ea42d20da39f31fd9e8831b\"" ], "inputs": { "codegen_source_sha256": "ef13d4bd541f6219f0dab7113e0e9c822c0f863335019544838d3cbfc4dfc653", "compile_flags": [ "-shared", "-fPIC", "-O3" ], "compiler": { "requested": "gcc", "resolved": "/usr/bin/x86_64-linux-gnu-gcc-13", "version": "gcc (Ubuntu 13.3.0-6ubuntu2~24.04.1) 13.3.0" }, "defines": { "CK_NUM_TOKENS": "8" }, "generated_source_sha256": "90f65f97cc873d7664a43d4f32ab705b2c6729417dc7d4a846107f54b422cef7", "include_dirs": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/include", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine" ], "kernel_library_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "libraries": [ "ckernel_engine", "m" ], "library_dirs": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build" ], "rpath": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build" }, "output_library": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain_stale_control.so", "output_library_sha256": "fa73ad4de91db054a895358fd852182a925e9dcb0f988c412bf097f7d39fc17e", "schema": "cke.v7.train_runtime_build_identity.v1" }, "substituted_library": { "bytes": 200184, "inode": 9437732, "mtime_ns": 1789978754408625566, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain_stale_control.so", "process_map_entries": [ "7fe8cd4d4000-7fe8cd4da000 r--p 00000000 08:01 9437732 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain_stale_control.so", "7fe8cd4da000-7fe8cd4f7000 r-xp 00006000 08:01 9437732 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain_stale_control.so", "7fe8cd4f7000-7fe8cd4fb000 r--p 00023000 08:01 9437732 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain_stale_control.so", "7fe8cd4fb000-7fe8cd4ff000 r--p 00027000 08:01 9437732 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain_stale_control.so", "7fe8cd4ff000-7fe8cd500000 rw-p 0002b000 08:01 9437732 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_certification/fp32_dense_2layer/libtrain_stale_control.so" ], "sha256": "fa73ad4de91db054a895358fd852182a925e9dcb0f988c412bf097f7d39fc17e" } } }, "failures": [], "current": true, "current_result": { "name": "v8 FP32 Generated Training Certification", "category": "training", "status": "pass", "duration_sec": 11.397255659103394, "stdout": "", "stderr": "", "error_msg": "", "perf_metric": null, "perf_unit": "", "baseline_perf": null, "perf_delta_pct": null, "sub_tests": [], "execution_kind": "make", "execution_id": "v8-training-certify-fp32", "execution_args": [] }, "note": "Current workflow evidence: version/v8/.cache/reports/training_certification_latest.json" }, "training_workflow": { "status": "pass", "passed": true, "current": true, "execution": { "actual_runtime_threads": 2, "cpu_affinity": [ 0, 1, 2, 3 ], "cpu_affinity_count": 4, "git_commit": "19694d9d9ba7cef51bce0e5ee7229041802bcb9e", "github": { "run_attempt": "1", "run_id": "35574289723", "sha": "19694d9d9ba7cef51bce0e5ee7229041802bcb9e" }, "hostname": "runnervmlun5p", "pid": 20638, "run_id": "28abfed5-b889-44fc-9a35-d10e12dd9075", "started_at": "2026-09-21T08:19:15.380879Z", "thread_environment": { "CK_NUM_THREADS": null, "OMP_NUM_THREADS": null } }, "configuration": { "architecture": "qwen3_style_dense_reduced", "beta1": 0.9, "beta2": 0.999, "d_model": 32, "dtype": "fp32", "epochs": 10, "eps": 1e-08, "grad_accum": 8, "heads": 4, "hidden": 64, "kv_heads": 4, "layers": 4, "lr": 0.0003, "numerical_contract": "strict_reference_order", "optimizer": "generated_c_adamw", "parallel_scaling_certified": false, "seq_len": 32, "token_presentations": 100000, "tokenizer": "byte", "training_config_sha256": "1bcbce6ba32093a655c122c1349cbde4e3c1bb8177525ee0175e2a2825735656", "unique_train_tokens": 10000, "vocab_size": 256, "weight_decay": 0.01 }, "checks": { "atomic_checkpoint_publication": { "passed": true, "power_loss_durability_claim": false }, "checkpoint_compatibility": { "authoritative_identity": "6b97bbcd00bef05b190c5a9035e20d07da575ed7a2a28d9f41473fb82237855e", "negative_controls": { "altered_cursor": { "authoritative_next_microstep": 1563, "passed": true }, "wrong_config": { "altered_field": "training_config_sha256", "passed": true }, "wrong_dataset": { "altered_field": "dataset_sha256", "passed": true }, "wrong_runtime": { "altered_field": "runtime_contract_sha256", "passed": true } }, "passed": true }, "final_partial_update": { "accum_counter": 0, "accum_tokens": 0, "contributing_tokens": 48, "counters_passed": true, "flush_return": 67, "gradient_max_abs_diff": 0.0038646459579467773, "moment_max_abs_diff": 0.0004012342542409897, "optimizer_step": 392, "passed": true, "present": true, "weight_max_abs_diff": 0.0024386756122112274 }, "fresh_process_resume": { "checkpoint_accum_counter": 3, "checkpoint_accum_tokens": 96, "checkpoint_owned_start_microstep": 1563, "optimizer_max_abs_diff": 0.0, "parent_pid": 20638, "passed": true, "weight_max_abs_diff": 0.0, "worker_pid": 20700 }, "inference_export": { "generation_logits_max_abs_diff": 0.0003528594970703125, "generation_sequence": [ 84, 104, 101, 32, 110, 101, 32, 97, 110, 32, 97, 110, 101, 32, 97, 110 ], "generation_steps": 12, "generation_tokens_match": true, "heldout_loss": 4.3848354367518985, "independent_training_build_directory": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference", "passed": true, "rebuilt_training_forward_logits_max_abs_diff": 0.0, "standalone_native_executable": "NOT_CERTIFIED", "v8_generated_source_sha256": "6b13d198f7f093b888a58470d3740f66c38487ee13246d79818c36c685e30ea8", "v8_inference_last_token_logits_max_abs_diff": 7.605552673339844e-05, "v8_inference_only_build_directory": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime", "v8_inference_tolerance": 0.001, "v8_library_sha256": "123b414016d149be78104139230be74a7877fddd0132f4fc76fd13dc28ce26ac" }, "learning": { "epochs": [ { "cke_mean_loss": 5.2214152694702145, "epoch": 1, "last_microstep": 313, "mean_loss": 5.2214152694702145, "pytorch_mean_loss": 5.221408633422851, "tokens": 10000 }, { "cke_mean_loss": 4.669050325775147, "epoch": 2, "last_microstep": 626, "mean_loss": 4.669050325775147, "pytorch_mean_loss": 4.669066997528076, "tokens": 10000 }, { "cke_mean_loss": 4.2020420753479, "epoch": 3, "last_microstep": 939, "mean_loss": 4.2020420753479, "pytorch_mean_loss": 4.202087162780762, "tokens": 10000 }, { "cke_mean_loss": 3.8673354217529297, "epoch": 4, "last_microstep": 1252, "mean_loss": 3.8673354217529297, "pytorch_mean_loss": 3.8673638538360597, "tokens": 10000 }, { "cke_mean_loss": 3.647550314331055, "epoch": 5, "last_microstep": 1565, "mean_loss": 3.647550314331055, "pytorch_mean_loss": 3.647552487564087, "tokens": 10000 }, { "cke_mean_loss": 3.503137083816528, "epoch": 6, "last_microstep": 1878, "mean_loss": 3.503137083816528, "pytorch_mean_loss": 3.5031209087371824, "tokens": 10000 }, { "cke_mean_loss": 3.398025325012207, "epoch": 7, "last_microstep": 2191, "mean_loss": 3.398025325012207, "pytorch_mean_loss": 3.3979940784454348, "tokens": 10000 }, { "cke_mean_loss": 3.3115394157409668, "epoch": 8, "last_microstep": 2504, "mean_loss": 3.3115394157409668, "pytorch_mean_loss": 3.3114934158325195, "tokens": 10000 }, { "cke_mean_loss": 3.2343138065338133, "epoch": 9, "last_microstep": 2817, "mean_loss": 3.2343138065338133, "pytorch_mean_loss": 3.234253077697754, "tokens": 10000 }, { "cke_mean_loss": 3.1565881313323976, "epoch": 10, "last_microstep": 3130, "mean_loss": 3.1565881313323976, "pytorch_mean_loss": 3.1565128894805907, "tokens": 10000 } ], "first_epoch_loss": 5.2214152694702145, "heldout_loss_after": 4.3848354367518985, "heldout_loss_before": 5.562910595507396, "last_epoch_loss": 3.1565881313323976, "passed": true, "sample_after": "The ne an ane an ane an ane an an a an a a an an an ", "sample_before": "The \ufffd\ufffd,\u02c9cQxE\ufffd\ufffdjxcN>\n\ufffd\ufffd\ufffdhO\u061e\u0014\u0003\ufffd\u0019x\ufffd'\ufffd\u0019x\ufffd\u0018,\u02c9cQw\ufffdcQw" }, "negative_control_detection": { "checkpoint_identity": { "altered_cursor": { "authoritative_next_microstep": 1563, "passed": true }, "wrong_config": { "altered_field": "training_config_sha256", "passed": true }, "wrong_dataset": { "altered_field": "dataset_sha256", "passed": true }, "wrong_runtime": { "altered_field": "runtime_contract_sha256", "passed": true } }, "final_partial_flush": { "applicable": true, "injected_lr": 0.03, "moment_delta_from_expected": 0.0, "passed": true, "weight_delta_from_expected": 0.07095326483249664 }, "passed": true }, "pytorch_trajectory": { "final_tensor_discrepancies": { "optimizer": [ { "max_abs_diff": 6.13778829574585e-05, "mean_abs_diff": 5.1254755817353725e-06, "name": "optimizer.m.final_ln_weight", "numel": 32, "reference_max_abs": 0.04555242508649826 }, { "max_abs_diff": 8.757924661040306e-06, "mean_abs_diff": 2.065472244794364e-06, "name": "optimizer.m.layer.0.b1", "numel": 128, "reference_max_abs": 0.002554186386987567 }, { "max_abs_diff": 0.0004012342542409897, "mean_abs_diff": 0.00012362371489871293, "name": "optimizer.m.layer.0.b2", "numel": 32, "reference_max_abs": 0.05907673016190529 }, { "max_abs_diff": 3.3705029636621475e-05, "mean_abs_diff": 1.1429556252551265e-05, "name": "optimizer.m.layer.0.bk", "numel": 32, "reference_max_abs": 0.004993596114218235 }, { "max_abs_diff": 0.00037310458719730377, "mean_abs_diff": 0.0001454479352105409, "name": "optimizer.m.layer.0.bo", "numel": 32, "reference_max_abs": 0.05942318215966225 }, { "max_abs_diff": 3.3770687878131866e-05, "mean_abs_diff": 1.4393874153029174e-05, "name": "optimizer.m.layer.0.bq", "numel": 32, "reference_max_abs": 0.006177674513310194 }, { "max_abs_diff": 6.437674164772034e-05, "mean_abs_diff": 1.6850139218149707e-05, "name": "optimizer.m.layer.0.bv", "numel": 32, "reference_max_abs": 0.010859882459044456 }, { "max_abs_diff": 9.308889275416732e-06, "mean_abs_diff": 4.738091774925124e-06, "name": "optimizer.m.layer.0.k_norm", "numel": 8, "reference_max_abs": 0.0016266326420009136 }, { "max_abs_diff": 8.352537406608462e-06, "mean_abs_diff": 3.001969616889255e-06, "name": "optimizer.m.layer.0.ln1_gamma", "numel": 32, "reference_max_abs": 0.0014161068247631192 }, { "max_abs_diff": 4.411820555105805e-06, "mean_abs_diff": 1.6220894849539036e-06, "name": "optimizer.m.layer.0.ln2_gamma", "numel": 32, "reference_max_abs": 0.0014566355384886265 }, { "max_abs_diff": 9.308147127740085e-06, "mean_abs_diff": 4.827758857572917e-06, "name": "optimizer.m.layer.0.q_norm", "numel": 8, "reference_max_abs": 0.001512247254140675 }, { "max_abs_diff": 2.0671868696808815e-05, "mean_abs_diff": 2.6351963242632337e-06, "name": "optimizer.m.layer.0.w1", "numel": 4096, "reference_max_abs": 0.005462396424263716 }, { "max_abs_diff": 3.748552990145981e-05, "mean_abs_diff": 3.74391493096482e-06, "name": "optimizer.m.layer.0.w2", "numel": 2048, "reference_max_abs": 0.007325064856559038 }, { "max_abs_diff": 0.00010278495028614998, "mean_abs_diff": 1.7642636521486565e-05, "name": "optimizer.m.layer.0.wk", "numel": 1024, "reference_max_abs": 0.018216704949736595 }, { "max_abs_diff": 5.1666516810655594e-05, "mean_abs_diff": 1.2185053492430598e-05, "name": "optimizer.m.layer.0.wo", "numel": 1024, "reference_max_abs": 0.010121841914951801 }, { "max_abs_diff": 6.955955177545547e-05, "mean_abs_diff": 1.3826544090989046e-05, "name": "optimizer.m.layer.0.wq", "numel": 1024, "reference_max_abs": 0.01074550487101078 }, { "max_abs_diff": 6.487482460215688e-05, "mean_abs_diff": 1.2625408999156207e-05, "name": "optimizer.m.layer.0.wv", "numel": 1024, "reference_max_abs": 0.00888671912252903 }, { "max_abs_diff": 4.3224135879427195e-06, "mean_abs_diff": 1.348751084151445e-06, "name": "optimizer.m.layer.1.b1", "numel": 128, "reference_max_abs": 0.0015871162759140134 }, { "max_abs_diff": 7.849931716918945e-05, "mean_abs_diff": 3.0159553716657683e-05, "name": "optimizer.m.layer.1.b2", "numel": 32, "reference_max_abs": 0.029450219124555588 }, { "max_abs_diff": 5.0108268624171615e-05, "mean_abs_diff": 8.42609370010905e-06, "name": "optimizer.m.layer.1.bk", "numel": 32, "reference_max_abs": 0.0026135300286114216 }, { "max_abs_diff": 0.00018821470439434052, "mean_abs_diff": 4.4635675294557586e-05, "name": "optimizer.m.layer.1.bo", "numel": 32, "reference_max_abs": 0.04515441134572029 }, { "max_abs_diff": 4.4282234739512205e-05, "mean_abs_diff": 1.3562387721322011e-05, "name": "optimizer.m.layer.1.bq", "numel": 32, "reference_max_abs": 0.003055550856515765 }, { "max_abs_diff": 2.8557726182043552e-05, "mean_abs_diff": 8.194628208002541e-06, "name": "optimizer.m.layer.1.bv", "numel": 32, "reference_max_abs": 0.006827468052506447 }, { "max_abs_diff": 2.966728061437607e-06, "mean_abs_diff": 1.8985579117725138e-06, "name": "optimizer.m.layer.1.k_norm", "numel": 8, "reference_max_abs": 0.0009987085359171033 }, { "max_abs_diff": 6.454312824644148e-06, "mean_abs_diff": 2.6132590846827952e-06, "name": "optimizer.m.layer.1.ln1_gamma", "numel": 32, "reference_max_abs": 0.0006189423729665577 }, { "max_abs_diff": 2.0521993064903654e-06, "mean_abs_diff": 8.500106218889414e-07, "name": "optimizer.m.layer.1.ln2_gamma", "numel": 32, "reference_max_abs": 0.0006688807625323534 }, { "max_abs_diff": 4.942266969010234e-06, "mean_abs_diff": 1.8020537027041428e-06, "name": "optimizer.m.layer.1.q_norm", "numel": 8, "reference_max_abs": 0.0007344809710048139 }, { "max_abs_diff": 1.6747042536735535e-05, "mean_abs_diff": 1.4978381841501687e-06, "name": "optimizer.m.layer.1.w1", "numel": 4096, "reference_max_abs": 0.0022200075909495354 }, { "max_abs_diff": 1.8115388229489326e-05, "mean_abs_diff": 1.7779361769498792e-06, "name": "optimizer.m.layer.1.w2", "numel": 2048, "reference_max_abs": 0.0029834068845957518 }, { "max_abs_diff": 6.659043719992042e-05, "mean_abs_diff": 1.0388325790700037e-05, "name": "optimizer.m.layer.1.wk", "numel": 1024, "reference_max_abs": 0.005724688060581684 }, { "max_abs_diff": 3.8541038520634174e-05, "mean_abs_diff": 7.531421488238266e-06, "name": "optimizer.m.layer.1.wo", "numel": 1024, "reference_max_abs": 0.005403027404099703 }, { "max_abs_diff": 7.408601231873035e-05, "mean_abs_diff": 1.2452188457245938e-05, "name": "optimizer.m.layer.1.wq", "numel": 1024, "reference_max_abs": 0.004823803901672363 }, { "max_abs_diff": 4.250218626111746e-05, "mean_abs_diff": 9.045837941812351e-06, "name": "optimizer.m.layer.1.wv", "numel": 1024, "reference_max_abs": 0.009451999329030514 }, { "max_abs_diff": 2.8440263122320175e-06, "mean_abs_diff": 7.048726047287346e-07, "name": "optimizer.m.layer.2.b1", "numel": 128, "reference_max_abs": 0.000839022162836045 }, { "max_abs_diff": 5.803816020488739e-05, "mean_abs_diff": 2.6768324460135773e-05, "name": "optimizer.m.layer.2.b2", "numel": 32, "reference_max_abs": 0.025311894714832306 }, { "max_abs_diff": 1.3754499377682805e-05, "mean_abs_diff": 3.811504484474426e-06, "name": "optimizer.m.layer.2.bk", "numel": 32, "reference_max_abs": 0.0008469597669318318 }, { "max_abs_diff": 7.965695112943649e-05, "mean_abs_diff": 2.8380149160511792e-05, "name": "optimizer.m.layer.2.bo", "numel": 32, "reference_max_abs": 0.02657248266041279 }, { "max_abs_diff": 2.082402352243662e-05, "mean_abs_diff": 4.897079634247348e-06, "name": "optimizer.m.layer.2.bq", "numel": 32, "reference_max_abs": 0.002063190331682563 }, { "max_abs_diff": 9.219205821864307e-06, "mean_abs_diff": 3.656578883237671e-06, "name": "optimizer.m.layer.2.bv", "numel": 32, "reference_max_abs": 0.0036350428126752377 }, { "max_abs_diff": 2.354237949475646e-06, "mean_abs_diff": 9.888907470667618e-07, "name": "optimizer.m.layer.2.k_norm", "numel": 8, "reference_max_abs": 0.0003624641103670001 }, { "max_abs_diff": 3.487046342343092e-06, "mean_abs_diff": 9.989848877012264e-07, "name": "optimizer.m.layer.2.ln1_gamma", "numel": 32, "reference_max_abs": 0.00044150566100142896 }, { "max_abs_diff": 1.828331733122468e-06, "mean_abs_diff": 4.7655191792728147e-07, "name": "optimizer.m.layer.2.ln2_gamma", "numel": 32, "reference_max_abs": 0.0004179107490926981 }, { "max_abs_diff": 3.522116458043456e-06, "mean_abs_diff": 1.024911966851505e-06, "name": "optimizer.m.layer.2.q_norm", "numel": 8, "reference_max_abs": 0.00043371861102059484 }, { "max_abs_diff": 8.42241570353508e-06, "mean_abs_diff": 8.226348313655762e-07, "name": "optimizer.m.layer.2.w1", "numel": 4096, "reference_max_abs": 0.0021707366686314344 }, { "max_abs_diff": 7.659924449399114e-06, "mean_abs_diff": 9.167695793621533e-07, "name": "optimizer.m.layer.2.w2", "numel": 2048, "reference_max_abs": 0.0021320979576557875 }, { "max_abs_diff": 2.806600423355121e-05, "mean_abs_diff": 4.320037533034338e-06, "name": "optimizer.m.layer.2.wk", "numel": 1024, "reference_max_abs": 0.0026555093936622143 }, { "max_abs_diff": 2.3291679099202156e-05, "mean_abs_diff": 3.6095939321967307e-06, "name": "optimizer.m.layer.2.wo", "numel": 1024, "reference_max_abs": 0.003130116034299135 }, { "max_abs_diff": 3.830069545074366e-05, "mean_abs_diff": 4.566591087495908e-06, "name": "optimizer.m.layer.2.wq", "numel": 1024, "reference_max_abs": 0.0028422835748642683 }, { "max_abs_diff": 2.3823929950594902e-05, "mean_abs_diff": 4.167106453678571e-06, "name": "optimizer.m.layer.2.wv", "numel": 1024, "reference_max_abs": 0.00487098703160882 }, { "max_abs_diff": 1.6425619833171368e-06, "mean_abs_diff": 3.2061120691651013e-07, "name": "optimizer.m.layer.3.b1", "numel": 128, "reference_max_abs": 0.0005140550783835351 }, { "max_abs_diff": 6.0528458561748266e-05, "mean_abs_diff": 2.5974515665438958e-05, "name": "optimizer.m.layer.3.b2", "numel": 32, "reference_max_abs": 0.025727784261107445 }, { "max_abs_diff": 4.512956365942955e-06, "mean_abs_diff": 1.4700985957460944e-06, "name": "optimizer.m.layer.3.bk", "numel": 32, "reference_max_abs": 0.0007776834536343813 }, { "max_abs_diff": 6.469659274443984e-05, "mean_abs_diff": 2.5824589101830497e-05, "name": "optimizer.m.layer.3.bo", "numel": 32, "reference_max_abs": 0.026388052850961685 }, { "max_abs_diff": 1.0373187251389027e-05, "mean_abs_diff": 2.5549882138875546e-06, "name": "optimizer.m.layer.3.bq", "numel": 32, "reference_max_abs": 0.0015243609668686986 }, { "max_abs_diff": 8.610542863607407e-06, "mean_abs_diff": 2.8499591735453578e-06, "name": "optimizer.m.layer.3.bv", "numel": 32, "reference_max_abs": 0.00208369642496109 }, { "max_abs_diff": 1.164109562523663e-06, "mean_abs_diff": 6.068873972253641e-07, "name": "optimizer.m.layer.3.k_norm", "numel": 8, "reference_max_abs": 0.00029260278097353876 }, { "max_abs_diff": 2.246335498057306e-06, "mean_abs_diff": 6.878341309857205e-07, "name": "optimizer.m.layer.3.ln1_gamma", "numel": 32, "reference_max_abs": 0.00040396227268502116 }, { "max_abs_diff": 8.181887096725404e-07, "mean_abs_diff": 1.6642179900827614e-07, "name": "optimizer.m.layer.3.ln2_gamma", "numel": 32, "reference_max_abs": 0.00020495473290793598 }, { "max_abs_diff": 1.2678210623562336e-06, "mean_abs_diff": 5.647580678669328e-07, "name": "optimizer.m.layer.3.q_norm", "numel": 8, "reference_max_abs": 0.0002858809020835906 }, { "max_abs_diff": 5.961832357570529e-06, "mean_abs_diff": 3.957686658395687e-07, "name": "optimizer.m.layer.3.w1", "numel": 4096, "reference_max_abs": 0.0018261770019307733 }, { "max_abs_diff": 5.900976248085499e-06, "mean_abs_diff": 4.866450922236254e-07, "name": "optimizer.m.layer.3.w2", "numel": 2048, "reference_max_abs": 0.002044906374067068 }, { "max_abs_diff": 1.9656290533021092e-05, "mean_abs_diff": 2.530407527956413e-06, "name": "optimizer.m.layer.3.wk", "numel": 1024, "reference_max_abs": 0.0031245131976902485 }, { "max_abs_diff": 1.861745840869844e-05, "mean_abs_diff": 3.0011665330675896e-06, "name": "optimizer.m.layer.3.wo", "numel": 1024, "reference_max_abs": 0.005968522746115923 }, { "max_abs_diff": 1.8987338989973068e-05, "mean_abs_diff": 2.8907315936521627e-06, "name": "optimizer.m.layer.3.wq", "numel": 1024, "reference_max_abs": 0.0030657993629574776 }, { "max_abs_diff": 2.0219711586833e-05, "mean_abs_diff": 3.051246267204988e-06, "name": "optimizer.m.layer.3.wv", "numel": 1024, "reference_max_abs": 0.004291176330298185 }, { "max_abs_diff": 0.00014741672202944756, "mean_abs_diff": 9.883415259537287e-07, "name": "optimizer.m.output.weight", "numel": 8192, "reference_max_abs": 0.1059625968337059 }, { "max_abs_diff": 0.00023396499454975128, "mean_abs_diff": 5.225725544732995e-06, "name": "optimizer.m.token_emb", "numel": 8192, "reference_max_abs": 0.047233615070581436 }, { "max_abs_diff": 2.2535095922648907e-07, "mean_abs_diff": 3.89777596865315e-08, "name": "optimizer.v.final_ln_weight", "numel": 32, "reference_max_abs": 0.0004140517266932875 }, { "max_abs_diff": 2.7126247914566193e-09, "mean_abs_diff": 4.123099639485872e-10, "name": "optimizer.v.layer.0.b1", "numel": 128, "reference_max_abs": 2.7893693186342716e-06 }, { "max_abs_diff": 5.338573828339577e-06, "mean_abs_diff": 1.3100252544973046e-06, "name": "optimizer.v.layer.0.b2", "numel": 32, "reference_max_abs": 0.0033436729572713375 }, { "max_abs_diff": 1.9294930098112673e-08, "mean_abs_diff": 7.47178141580207e-09, "name": "optimizer.v.layer.0.bk", "numel": 32, "reference_max_abs": 4.348440597823355e-06 }, { "max_abs_diff": 5.992129445075989e-06, "mean_abs_diff": 1.6008925740607083e-06, "name": "optimizer.v.layer.0.bo", "numel": 32, "reference_max_abs": 0.00432991236448288 }, { "max_abs_diff": 6.384561856975779e-08, "mean_abs_diff": 1.0460659893851698e-08, "name": "optimizer.v.layer.0.bq", "numel": 32, "reference_max_abs": 8.692153642186895e-06 }, { "max_abs_diff": 2.2782842279411852e-07, "mean_abs_diff": 4.5308922835829435e-08, "name": "optimizer.v.layer.0.bv", "numel": 32, "reference_max_abs": 6.904634210513905e-05 }, { "max_abs_diff": 2.416726374576683e-09, "mean_abs_diff": 7.512301891665629e-10, "name": "optimizer.v.layer.0.k_norm", "numel": 8, "reference_max_abs": 6.169541961753566e-07 }, { "max_abs_diff": 1.6312355910486076e-09, "mean_abs_diff": 6.183793477276822e-10, "name": "optimizer.v.layer.0.ln1_gamma", "numel": 32, "reference_max_abs": 5.603984050139843e-07 }, { "max_abs_diff": 5.153992788109463e-10, "mean_abs_diff": 1.4155057681541905e-10, "name": "optimizer.v.layer.0.ln2_gamma", "numel": 32, "reference_max_abs": 5.504492150976148e-07 }, { "max_abs_diff": 2.4229507289419416e-09, "mean_abs_diff": 7.620073461112042e-10, "name": "optimizer.v.layer.0.q_norm", "numel": 8, "reference_max_abs": 4.72709245968872e-07 }, { "max_abs_diff": 1.1941210686927661e-08, "mean_abs_diff": 4.815688958714759e-10, "name": "optimizer.v.layer.0.w1", "numel": 4096, "reference_max_abs": 7.71555096434895e-06 }, { "max_abs_diff": 4.7750745579833165e-08, "mean_abs_diff": 1.1928026344421028e-09, "name": "optimizer.v.layer.0.w2", "numel": 2048, "reference_max_abs": 8.919327228795737e-06 }, { "max_abs_diff": 3.2979005482047796e-07, "mean_abs_diff": 1.7933954765680937e-08, "name": "optimizer.v.layer.0.wk", "numel": 1024, "reference_max_abs": 6.891953671583906e-05 }, { "max_abs_diff": 6.276422936934978e-08, "mean_abs_diff": 8.204794177402164e-09, "name": "optimizer.v.layer.0.wo", "numel": 1024, "reference_max_abs": 2.123128615494352e-05 }, { "max_abs_diff": 9.58652890403755e-08, "mean_abs_diff": 1.26589823068457e-08, "name": "optimizer.v.layer.0.wq", "numel": 1024, "reference_max_abs": 1.5190847079793457e-05 }, { "max_abs_diff": 1.5910336514934897e-07, "mean_abs_diff": 1.146525008266508e-08, "name": "optimizer.v.layer.0.wv", "numel": 1024, "reference_max_abs": 5.092455103294924e-05 }, { "max_abs_diff": 2.5716531126818154e-09, "mean_abs_diff": 2.4386231922690627e-10, "name": "optimizer.v.layer.1.b1", "numel": 128, "reference_max_abs": 1.8507369077269686e-06 }, { "max_abs_diff": 2.027139998972416e-06, "mean_abs_diff": 4.070147952006664e-07, "name": "optimizer.v.layer.1.b2", "numel": 32, "reference_max_abs": 0.0012642254587262869 }, { "max_abs_diff": 1.005298599920934e-08, "mean_abs_diff": 3.700947548423983e-09, "name": "optimizer.v.layer.1.bk", "numel": 32, "reference_max_abs": 1.6373633116018027e-06 }, { "max_abs_diff": 3.066263161599636e-06, "mean_abs_diff": 6.16538727626903e-07, "name": "optimizer.v.layer.1.bo", "numel": 32, "reference_max_abs": 0.0014783082297071815 }, { "max_abs_diff": 2.714705260586925e-08, "mean_abs_diff": 8.052777999978389e-09, "name": "optimizer.v.layer.1.bq", "numel": 32, "reference_max_abs": 5.817753844894469e-06 }, { "max_abs_diff": 8.923598215915263e-08, "mean_abs_diff": 1.9245170790327393e-08, "name": "optimizer.v.layer.1.bv", "numel": 32, "reference_max_abs": 2.979060445795767e-05 }, { "max_abs_diff": 6.461675639002351e-10, "mean_abs_diff": 3.206634957564347e-10, "name": "optimizer.v.layer.1.k_norm", "numel": 8, "reference_max_abs": 1.5427259825173678e-07 }, { "max_abs_diff": 1.8109176380676217e-09, "mean_abs_diff": 3.023476469188324e-10, "name": "optimizer.v.layer.1.ln1_gamma", "numel": 32, "reference_max_abs": 2.980892759296694e-07 }, { "max_abs_diff": 8.936353879107628e-10, "mean_abs_diff": 1.0136610995026274e-10, "name": "optimizer.v.layer.1.ln2_gamma", "numel": 32, "reference_max_abs": 3.691318681831035e-07 }, { "max_abs_diff": 7.381473210443801e-10, "mean_abs_diff": 3.3087133033404825e-10, "name": "optimizer.v.layer.1.q_norm", "numel": 8, "reference_max_abs": 1.6441877903616842e-07 }, { "max_abs_diff": 9.570840120431967e-09, "mean_abs_diff": 2.613114447047593e-10, "name": "optimizer.v.layer.1.w1", "numel": 4096, "reference_max_abs": 6.854266757727601e-06 }, { "max_abs_diff": 1.7097590898629278e-08, "mean_abs_diff": 5.779593470478517e-10, "name": "optimizer.v.layer.1.w2", "numel": 2048, "reference_max_abs": 2.083455001411494e-05 }, { "max_abs_diff": 6.108621164457873e-08, "mean_abs_diff": 7.06130087735346e-09, "name": "optimizer.v.layer.1.wk", "numel": 1024, "reference_max_abs": 8.554938176530413e-06 }, { "max_abs_diff": 5.9411831898614764e-08, "mean_abs_diff": 3.998587239095741e-09, "name": "optimizer.v.layer.1.wo", "numel": 1024, "reference_max_abs": 1.5220134628179949e-05 }, { "max_abs_diff": 6.977461453061551e-08, "mean_abs_diff": 7.172159310897541e-09, "name": "optimizer.v.layer.1.wq", "numel": 1024, "reference_max_abs": 1.5122373042686377e-05 }, { "max_abs_diff": 1.2698001228272915e-07, "mean_abs_diff": 7.2237624770821185e-09, "name": "optimizer.v.layer.1.wv", "numel": 1024, "reference_max_abs": 3.483444743324071e-05 }, { "max_abs_diff": 2.3320581021835096e-09, "mean_abs_diff": 1.3172210544531993e-10, "name": "optimizer.v.layer.2.b1", "numel": 128, "reference_max_abs": 1.1324217439323547e-06 }, { "max_abs_diff": 8.05361196398735e-07, "mean_abs_diff": 1.454827724955976e-07, "name": "optimizer.v.layer.2.b2", "numel": 32, "reference_max_abs": 0.001089070807211101 }, { "max_abs_diff": 3.3890330541908043e-09, "mean_abs_diff": 8.31716362448276e-10, "name": "optimizer.v.layer.2.bk", "numel": 32, "reference_max_abs": 3.9583639477314136e-07 }, { "max_abs_diff": 1.3400858733803034e-06, "mean_abs_diff": 2.2626909412792884e-07, "name": "optimizer.v.layer.2.bo", "numel": 32, "reference_max_abs": 0.0011309873079881072 }, { "max_abs_diff": 1.140369931817986e-08, "mean_abs_diff": 1.2460295018001943e-09, "name": "optimizer.v.layer.2.bq", "numel": 32, "reference_max_abs": 1.4896837683409103e-06 }, { "max_abs_diff": 3.613240551203489e-08, "mean_abs_diff": 7.160313231224791e-09, "name": "optimizer.v.layer.2.bv", "numel": 32, "reference_max_abs": 1.6938818589551374e-05 }, { "max_abs_diff": 1.2358114531707542e-10, "mean_abs_diff": 6.241468453183074e-11, "name": "optimizer.v.layer.2.k_norm", "numel": 8, "reference_max_abs": 7.883939900921177e-08 }, { "max_abs_diff": 3.119566827081144e-10, "mean_abs_diff": 7.487765962821413e-11, "name": "optimizer.v.layer.2.ln1_gamma", "numel": 32, "reference_max_abs": 2.4813235199871997e-07 }, { "max_abs_diff": 3.865707753902825e-10, "mean_abs_diff": 4.2644734965513464e-11, "name": "optimizer.v.layer.2.ln2_gamma", "numel": 32, "reference_max_abs": 8.716418165022333e-08 }, { "max_abs_diff": 2.78163270195364e-10, "mean_abs_diff": 7.817640979013163e-11, "name": "optimizer.v.layer.2.q_norm", "numel": 8, "reference_max_abs": 7.888771591524346e-08 }, { "max_abs_diff": 7.3960109148174524e-09, "mean_abs_diff": 1.3837880841194306e-10, "name": "optimizer.v.layer.2.w1", "numel": 4096, "reference_max_abs": 2.487690835550893e-06 }, { "max_abs_diff": 5.100503130961442e-09, "mean_abs_diff": 1.6566852056865145e-10, "name": "optimizer.v.layer.2.w2", "numel": 2048, "reference_max_abs": 1.4863742308079964e-06 }, { "max_abs_diff": 1.1335941962897778e-08, "mean_abs_diff": 1.3912645480118613e-09, "name": "optimizer.v.layer.2.wk", "numel": 1024, "reference_max_abs": 3.1675815534981666e-06 }, { "max_abs_diff": 2.8609974833671004e-08, "mean_abs_diff": 1.9118775451687497e-09, "name": "optimizer.v.layer.2.wo", "numel": 1024, "reference_max_abs": 1.5684274330851622e-05 }, { "max_abs_diff": 1.989985776162939e-08, "mean_abs_diff": 1.0173155651216348e-09, "name": "optimizer.v.layer.2.wq", "numel": 1024, "reference_max_abs": 2.4884714093786897e-06 }, { "max_abs_diff": 5.224592314334586e-08, "mean_abs_diff": 3.1841747016869704e-09, "name": "optimizer.v.layer.2.wv", "numel": 1024, "reference_max_abs": 1.5635136151104234e-05 }, { "max_abs_diff": 1.5666614672227297e-09, "mean_abs_diff": 3.9650692451598246e-11, "name": "optimizer.v.layer.3.b1", "numel": 128, "reference_max_abs": 6.492778084066231e-07 }, { "max_abs_diff": 1.2670352589339018e-06, "mean_abs_diff": 1.3889894034946337e-07, "name": "optimizer.v.layer.3.b2", "numel": 32, "reference_max_abs": 0.0007301259902305901 }, { "max_abs_diff": 1.3410215160547523e-09, "mean_abs_diff": 2.561724443683744e-10, "name": "optimizer.v.layer.3.bk", "numel": 32, "reference_max_abs": 2.4867227921276935e-07 }, { "max_abs_diff": 1.4163379091769457e-06, "mean_abs_diff": 1.4565875972039066e-07, "name": "optimizer.v.layer.3.bo", "numel": 32, "reference_max_abs": 0.0007472163997590542 }, { "max_abs_diff": 3.3417109079891816e-09, "mean_abs_diff": 7.738160112680248e-10, "name": "optimizer.v.layer.3.bq", "numel": 32, "reference_max_abs": 1.6519503560630255e-06 }, { "max_abs_diff": 9.27138898987323e-09, "mean_abs_diff": 1.7390497930591664e-09, "name": "optimizer.v.layer.3.bv", "numel": 32, "reference_max_abs": 1.991740100493189e-05 }, { "max_abs_diff": 2.695585976653092e-10, "mean_abs_diff": 4.9879766983451646e-11, "name": "optimizer.v.layer.3.k_norm", "numel": 8, "reference_max_abs": 8.465558920534022e-08 }, { "max_abs_diff": 1.8813750557455933e-10, "mean_abs_diff": 4.390127150699641e-11, "name": "optimizer.v.layer.3.ln1_gamma", "numel": 32, "reference_max_abs": 1.514763425802812e-07 }, { "max_abs_diff": 3.569100570643968e-11, "mean_abs_diff": 4.949809659371418e-12, "name": "optimizer.v.layer.3.ln2_gamma", "numel": 32, "reference_max_abs": 1.7492263637564065e-08 }, { "max_abs_diff": 2.6931701313515077e-10, "mean_abs_diff": 4.550831933514132e-11, "name": "optimizer.v.layer.3.q_norm", "numel": 8, "reference_max_abs": 8.461091027811563e-08 }, { "max_abs_diff": 2.8724116418743506e-09, "mean_abs_diff": 3.864134359710114e-11, "name": "optimizer.v.layer.3.w1", "numel": 4096, "reference_max_abs": 1.2544201126729604e-06 }, { "max_abs_diff": 7.12918790668482e-09, "mean_abs_diff": 1.1262642207743312e-10, "name": "optimizer.v.layer.3.w2", "numel": 2048, "reference_max_abs": 3.1106667393032694e-06 }, { "max_abs_diff": 1.1054225979023613e-08, "mean_abs_diff": 6.389954121388541e-10, "name": "optimizer.v.layer.3.wk", "numel": 1024, "reference_max_abs": 2.6132879611395765e-06 }, { "max_abs_diff": 7.409653335344046e-08, "mean_abs_diff": 2.7451634299069383e-09, "name": "optimizer.v.layer.3.wo", "numel": 1024, "reference_max_abs": 4.691631329478696e-05 }, { "max_abs_diff": 1.1179054126841947e-08, "mean_abs_diff": 8.126422867960059e-10, "name": "optimizer.v.layer.3.wq", "numel": 1024, "reference_max_abs": 3.4429274364811135e-06 }, { "max_abs_diff": 2.5679128157207742e-08, "mean_abs_diff": 1.7146444264426464e-09, "name": "optimizer.v.layer.3.wv", "numel": 1024, "reference_max_abs": 2.4467117327731103e-05 }, { "max_abs_diff": 4.4800108298659325e-06, "mean_abs_diff": 4.789165952701069e-09, "name": "optimizer.v.output.weight", "numel": 8192, "reference_max_abs": 0.004440011456608772 }, { "max_abs_diff": 3.0489754863083363e-06, "mean_abs_diff": 1.516946035451383e-08, "name": "optimizer.v.token_emb", "numel": 8192, "reference_max_abs": 0.000604337255936116 } ], "partial_window_gradients": [ { "max_abs_diff": 4.15164977312088e-05, "mean_abs_diff": 1.2086980859749019e-05, "name": "final_ln_weight", "numel": 32, "reference_max_abs": 0.040681932121515274 }, { "max_abs_diff": 8.932780474424362e-05, "mean_abs_diff": 1.7507796655991115e-05, "name": "layer.0.b1", "numel": 128, "reference_max_abs": 0.01556966733187437 }, { "max_abs_diff": 0.002591364085674286, "mean_abs_diff": 0.0008927698363550007, "name": "layer.0.b2", "numel": 32, "reference_max_abs": 0.5431657433509827 }, { "max_abs_diff": 0.0003114696592092514, "mean_abs_diff": 0.00011620884470175952, "name": "layer.0.bk", "numel": 32, "reference_max_abs": 0.042613327503204346 }, { "max_abs_diff": 0.0038646459579467773, "mean_abs_diff": 0.0010468473192304373, "name": "layer.0.bo", "numel": 32, "reference_max_abs": 0.6464266180992126 }, { "max_abs_diff": 0.00037060491740703583, "mean_abs_diff": 0.00010179997480008751, "name": "layer.0.bq", "numel": 32, "reference_max_abs": 0.0605577789247036 }, { "max_abs_diff": 0.0007758364081382751, "mean_abs_diff": 0.00018089768127538264, "name": "layer.0.bv", "numel": 32, "reference_max_abs": 0.11701565980911255 }, { "max_abs_diff": 4.820665344595909e-05, "mean_abs_diff": 3.478761209407821e-05, "name": "layer.0.k_norm", "numel": 8, "reference_max_abs": 0.012150316499173641 }, { "max_abs_diff": 9.697279892861843e-05, "mean_abs_diff": 2.686345760594122e-05, "name": "layer.0.ln1_gamma", "numel": 32, "reference_max_abs": 0.013802426867187023 }, { "max_abs_diff": 3.1284638680517673e-05, "mean_abs_diff": 1.0062567525892518e-05, "name": "layer.0.ln2_gamma", "numel": 32, "reference_max_abs": 0.010509413667023182 }, { "max_abs_diff": 6.0907972510904074e-05, "mean_abs_diff": 4.459646515897475e-05, "name": "layer.0.q_norm", "numel": 8, "reference_max_abs": 0.009219720959663391 }, { "max_abs_diff": 0.00017104856669902802, "mean_abs_diff": 2.0810140995308757e-05, "name": "layer.0.w1", "numel": 4096, "reference_max_abs": 0.03250782564282417 }, { "max_abs_diff": 0.0004224628210067749, "mean_abs_diff": 2.5466244551353157e-05, "name": "layer.0.w2", "numel": 2048, "reference_max_abs": 0.044930990785360336 }, { "max_abs_diff": 0.0008071176707744598, "mean_abs_diff": 0.00014859024668112397, "name": "layer.0.wk", "numel": 1024, "reference_max_abs": 0.21794478595256805 }, { "max_abs_diff": 0.0005457159131765366, "mean_abs_diff": 9.962912008631974e-05, "name": "layer.0.wo", "numel": 1024, "reference_max_abs": 0.0852627083659172 }, { "max_abs_diff": 0.0007025599479675293, "mean_abs_diff": 0.00010902604117291048, "name": "layer.0.wq", "numel": 1024, "reference_max_abs": 0.10185468196868896 }, { "max_abs_diff": 0.0005569364875555038, "mean_abs_diff": 0.00010072087752632797, "name": "layer.0.wv", "numel": 1024, "reference_max_abs": 0.13255788385868073 }, { "max_abs_diff": 4.891751450486481e-05, "mean_abs_diff": 8.994482413982041e-06, "name": "layer.1.b1", "numel": 128, "reference_max_abs": 0.01715818978846073 }, { "max_abs_diff": 0.0009372271597385406, "mean_abs_diff": 0.0003410645294934511, "name": "layer.1.b2", "numel": 32, "reference_max_abs": 0.21124184131622314 }, { "max_abs_diff": 0.00029055122286081314, "mean_abs_diff": 7.091982843121514e-05, "name": "layer.1.bk", "numel": 32, "reference_max_abs": 0.01514804270118475 }, { "max_abs_diff": 0.0010353773832321167, "mean_abs_diff": 0.00043161638313904405, "name": "layer.1.bo", "numel": 32, "reference_max_abs": 0.3128746449947357 }, { "max_abs_diff": 0.0002861917018890381, "mean_abs_diff": 9.591868001734838e-05, "name": "layer.1.bq", "numel": 32, "reference_max_abs": 0.03295781835913658 }, { "max_abs_diff": 0.0002559833228588104, "mean_abs_diff": 7.64585129218176e-05, "name": "layer.1.bv", "numel": 32, "reference_max_abs": 0.06841766834259033 }, { "max_abs_diff": 7.712794467806816e-05, "mean_abs_diff": 2.358978963457048e-05, "name": "layer.1.k_norm", "numel": 8, "reference_max_abs": 0.005984985735267401 }, { "max_abs_diff": 5.6809745728969574e-05, "mean_abs_diff": 2.0804225641768426e-05, "name": "layer.1.ln1_gamma", "numel": 32, "reference_max_abs": 0.006969836074858904 }, { "max_abs_diff": 1.733240787871182e-05, "mean_abs_diff": 6.18735884927446e-06, "name": "layer.1.ln2_gamma", "numel": 32, "reference_max_abs": 0.0033943888265639544 }, { "max_abs_diff": 6.531376857310534e-05, "mean_abs_diff": 2.1469993953360245e-05, "name": "layer.1.q_norm", "numel": 8, "reference_max_abs": 0.005985770374536514 }, { "max_abs_diff": 9.141117334365845e-05, "mean_abs_diff": 1.083596725948155e-05, "name": "layer.1.w1", "numel": 4096, "reference_max_abs": 0.026976674795150757 }, { "max_abs_diff": 0.00015805335715413094, "mean_abs_diff": 1.367651202599518e-05, "name": "layer.1.w2", "numel": 2048, "reference_max_abs": 0.03411191701889038 }, { "max_abs_diff": 0.0004842914640903473, "mean_abs_diff": 8.466669532936066e-05, "name": "layer.1.wk", "numel": 1024, "reference_max_abs": 0.040902089327573776 }, { "max_abs_diff": 0.0003304830752313137, "mean_abs_diff": 6.047444912837818e-05, "name": "layer.1.wo", "numel": 1024, "reference_max_abs": 0.03869878873229027 }, { "max_abs_diff": 0.0005069458857178688, "mean_abs_diff": 9.939067240338773e-05, "name": "layer.1.wq", "numel": 1024, "reference_max_abs": 0.041818637400865555 }, { "max_abs_diff": 0.00048487819731235504, "mean_abs_diff": 7.021998317213729e-05, "name": "layer.1.wv", "numel": 1024, "reference_max_abs": 0.07763058692216873 }, { "max_abs_diff": 4.1703926399350166e-05, "mean_abs_diff": 4.7110174818953965e-06, "name": "layer.2.b1", "numel": 128, "reference_max_abs": 0.006375929806381464 }, { "max_abs_diff": 0.0005414634943008423, "mean_abs_diff": 0.0002420655218884349, "name": "layer.2.b2", "numel": 32, "reference_max_abs": 0.20622999966144562 }, { "max_abs_diff": 9.68927051872015e-05, "mean_abs_diff": 3.4445663914084435e-05, "name": "layer.2.bk", "numel": 32, "reference_max_abs": 0.008237677626311779 }, { "max_abs_diff": 0.0005987199838273227, "mean_abs_diff": 0.00027328860596753657, "name": "layer.2.bo", "numel": 32, "reference_max_abs": 0.2271944284439087 }, { "max_abs_diff": 0.00016331393271684647, "mean_abs_diff": 4.185899888398126e-05, "name": "layer.2.bq", "numel": 32, "reference_max_abs": 0.015411165542900562 }, { "max_abs_diff": 0.00010353606194257736, "mean_abs_diff": 3.477161590126343e-05, "name": "layer.2.bv", "numel": 32, "reference_max_abs": 0.032063886523246765 }, { "max_abs_diff": 1.221813727170229e-05, "mean_abs_diff": 7.227274181786925e-06, "name": "layer.2.k_norm", "numel": 8, "reference_max_abs": 0.00158941640984267 }, { "max_abs_diff": 2.850964665412903e-05, "mean_abs_diff": 8.582117516198196e-06, "name": "layer.2.ln1_gamma", "numel": 32, "reference_max_abs": 0.0026582174468785524 }, { "max_abs_diff": 1.0013871360570192e-05, "mean_abs_diff": 3.3313592666672776e-06, "name": "layer.2.ln2_gamma", "numel": 32, "reference_max_abs": 0.001103983260691166 }, { "max_abs_diff": 2.77648214250803e-05, "mean_abs_diff": 9.327726729679853e-06, "name": "layer.2.q_norm", "numel": 8, "reference_max_abs": 0.0021051682997494936 }, { "max_abs_diff": 5.454523488879204e-05, "mean_abs_diff": 5.899318239244167e-06, "name": "layer.2.w1", "numel": 4096, "reference_max_abs": 0.011454747058451176 }, { "max_abs_diff": 7.439218461513519e-05, "mean_abs_diff": 6.517922429338796e-06, "name": "layer.2.w2", "numel": 2048, "reference_max_abs": 0.01065365131944418 }, { "max_abs_diff": 0.0002978779375553131, "mean_abs_diff": 3.667527198558673e-05, "name": "layer.2.wk", "numel": 1024, "reference_max_abs": 0.020030086860060692 }, { "max_abs_diff": 0.00023303832858800888, "mean_abs_diff": 2.8778955311281607e-05, "name": "layer.2.wo", "numel": 1024, "reference_max_abs": 0.027800867334008217 }, { "max_abs_diff": 0.00034144381061196327, "mean_abs_diff": 3.9775164623279124e-05, "name": "layer.2.wq", "numel": 1024, "reference_max_abs": 0.021691134199500084 }, { "max_abs_diff": 0.00021859165281057358, "mean_abs_diff": 3.356077286298387e-05, "name": "layer.2.wv", "numel": 1024, "reference_max_abs": 0.04800127446651459 }, { "max_abs_diff": 1.4773395378142595e-05, "mean_abs_diff": 2.4025925995374564e-06, "name": "layer.3.b1", "numel": 128, "reference_max_abs": 0.0032030486036092043 }, { "max_abs_diff": 0.00035553425550460815, "mean_abs_diff": 0.00015911336231511086, "name": "layer.3.b2", "numel": 32, "reference_max_abs": 0.19603633880615234 }, { "max_abs_diff": 5.6267017498612404e-05, "mean_abs_diff": 1.6315652828780003e-05, "name": "layer.3.bk", "numel": 32, "reference_max_abs": 0.00343754212372005 }, { "max_abs_diff": 0.00037219375371932983, "mean_abs_diff": 0.00016485519881825894, "name": "layer.3.bo", "numel": 32, "reference_max_abs": 0.19791246950626373 }, { "max_abs_diff": 7.035466842353344e-05, "mean_abs_diff": 1.9899758626706898e-05, "name": "layer.3.bq", "numel": 32, "reference_max_abs": 0.008718959055840969 }, { "max_abs_diff": 5.7967379689216614e-05, "mean_abs_diff": 2.5403514882782474e-05, "name": "layer.3.bv", "numel": 32, "reference_max_abs": 0.01847936026751995 }, { "max_abs_diff": 8.886214345693588e-06, "mean_abs_diff": 4.587991497828625e-06, "name": "layer.3.k_norm", "numel": 8, "reference_max_abs": 0.001958954380825162 }, { "max_abs_diff": 1.0446907253935933e-05, "mean_abs_diff": 4.437180450622691e-06, "name": "layer.3.ln1_gamma", "numel": 32, "reference_max_abs": 0.0030768278520554304 }, { "max_abs_diff": 2.771559593384154e-06, "mean_abs_diff": 9.083006489163381e-07, "name": "layer.3.ln2_gamma", "numel": 32, "reference_max_abs": 0.0009500349988229573 }, { "max_abs_diff": 9.864772437140346e-06, "mean_abs_diff": 4.647601599572226e-06, "name": "layer.3.q_norm", "numel": 8, "reference_max_abs": 0.0017992965877056122 }, { "max_abs_diff": 2.740207128226757e-05, "mean_abs_diff": 2.840695742634125e-06, "name": "layer.3.w1", "numel": 4096, "reference_max_abs": 0.01071630883961916 }, { "max_abs_diff": 3.5801902413368225e-05, "mean_abs_diff": 3.2391524200647837e-06, "name": "layer.3.w2", "numel": 2048, "reference_max_abs": 0.013236235827207565 }, { "max_abs_diff": 0.00016044871881604195, "mean_abs_diff": 2.0486597350100055e-05, "name": "layer.3.wk", "numel": 1024, "reference_max_abs": 0.017910419031977654 }, { "max_abs_diff": 0.00012141384650021791, "mean_abs_diff": 2.019587736867834e-05, "name": "layer.3.wo", "numel": 1024, "reference_max_abs": 0.04900243878364563 }, { "max_abs_diff": 0.0001720758154988289, "mean_abs_diff": 2.3674459953326732e-05, "name": "layer.3.wq", "numel": 1024, "reference_max_abs": 0.02217108942568302 }, { "max_abs_diff": 0.00013742782175540924, "mean_abs_diff": 2.224378113169223e-05, "name": "layer.3.wv", "numel": 1024, "reference_max_abs": 0.040741272270679474 }, { "max_abs_diff": 0.0002980455756187439, "mean_abs_diff": 4.652274128602585e-06, "name": "output.weight", "numel": 8192, "reference_max_abs": 0.1727251261472702 }, { "max_abs_diff": 0.0024294257164001465, "mean_abs_diff": 3.257976277382113e-05, "name": "token_emb", "numel": 8192, "reference_max_abs": 0.5125349164009094 } ], "weights": [ { "max_abs_diff": 0.00018477439880371094, "mean_abs_diff": 1.3221055269241333e-05, "name": "final_ln_weight", "numel": 32, "reference_max_abs": 1.1587904691696167 }, { "max_abs_diff": 9.561137994751334e-05, "mean_abs_diff": 1.8457527403370477e-05, "name": "layer.0.b1", "numel": 128, "reference_max_abs": 0.033238135278224945 }, { "max_abs_diff": 1.8004560843110085e-05, "mean_abs_diff": 3.5427533475740347e-06, "name": "layer.0.b2", "numel": 32, "reference_max_abs": 0.005040556658059359 }, { "max_abs_diff": 9.097688598558307e-05, "mean_abs_diff": 2.9090093448758125e-05, "name": "layer.0.bk", "numel": 32, "reference_max_abs": 0.03264597803354263 }, { "max_abs_diff": 1.4569610357284546e-05, "mean_abs_diff": 4.020038431917783e-06, "name": "layer.0.bo", "numel": 32, "reference_max_abs": 0.005147565621882677 }, { "max_abs_diff": 9.316205978393555e-05, "mean_abs_diff": 2.9597831598948687e-05, "name": "layer.0.bq", "numel": 32, "reference_max_abs": 0.023619789630174637 }, { "max_abs_diff": 1.7460784874856472e-05, "mean_abs_diff": 4.572934813040774e-06, "name": "layer.0.bv", "numel": 32, "reference_max_abs": 0.007215855177491903 }, { "max_abs_diff": 7.009506225585938e-05, "mean_abs_diff": 4.139542579650879e-05, "name": "layer.0.k_norm", "numel": 8, "reference_max_abs": 1.0815914869308472 }, { "max_abs_diff": 0.00010657310485839844, "mean_abs_diff": 3.0566006898880005e-05, "name": "layer.0.ln1_gamma", "numel": 32, "reference_max_abs": 1.0288059711456299 }, { "max_abs_diff": 5.9604644775390625e-05, "mean_abs_diff": 1.595914363861084e-05, "name": "layer.0.ln2_gamma", "numel": 32, "reference_max_abs": 1.078316569328308 }, { "max_abs_diff": 7.30752944946289e-05, "mean_abs_diff": 4.5418739318847656e-05, "name": "layer.0.q_norm", "numel": 8, "reference_max_abs": 1.07564377784729 }, { "max_abs_diff": 0.00022746995091438293, "mean_abs_diff": 3.147849565721117e-05, "name": "layer.0.w1", "numel": 4096, "reference_max_abs": 0.10106683522462845 }, { "max_abs_diff": 0.0002300255000591278, "mean_abs_diff": 2.9381100830505602e-05, "name": "layer.0.w2", "numel": 2048, "reference_max_abs": 0.1230408325791359 }, { "max_abs_diff": 0.0001901155337691307, "mean_abs_diff": 3.596942042349838e-05, "name": "layer.0.wk", "numel": 1024, "reference_max_abs": 0.09813330322504044 }, { "max_abs_diff": 0.00019652210175991058, "mean_abs_diff": 2.4872668291209266e-05, "name": "layer.0.wo", "numel": 1024, "reference_max_abs": 0.06762653589248657 }, { "max_abs_diff": 0.00044925231486558914, "mean_abs_diff": 3.435065445955843e-05, "name": "layer.0.wq", "numel": 1024, "reference_max_abs": 0.0848502367734909 }, { "max_abs_diff": 0.0006211176514625549, "mean_abs_diff": 2.2850341338198632e-05, "name": "layer.0.wv", "numel": 1024, "reference_max_abs": 0.08720789849758148 }, { "max_abs_diff": 0.00011343788355588913, "mean_abs_diff": 1.536999479867518e-05, "name": "layer.1.b1", "numel": 128, "reference_max_abs": 0.047799497842788696 }, { "max_abs_diff": 5.2300747483968735e-06, "mean_abs_diff": 2.0449820112844463e-06, "name": "layer.1.b2", "numel": 32, "reference_max_abs": 0.007889511063694954 }, { "max_abs_diff": 0.0002818913199007511, "mean_abs_diff": 4.643788270186633e-05, "name": "layer.1.bk", "numel": 32, "reference_max_abs": 0.02141890488564968 }, { "max_abs_diff": 6.382993888109922e-06, "mean_abs_diff": 2.441381184326019e-06, "name": "layer.1.bo", "numel": 32, "reference_max_abs": 0.007001741789281368 }, { "max_abs_diff": 8.68113711476326e-05, "mean_abs_diff": 2.9480041121132672e-05, "name": "layer.1.bq", "numel": 32, "reference_max_abs": 0.02565094083547592 }, { "max_abs_diff": 1.1422904208302498e-05, "mean_abs_diff": 3.1997506084735505e-06, "name": "layer.1.bv", "numel": 32, "reference_max_abs": 0.006465289741754532 }, { "max_abs_diff": 0.00010406970977783203, "mean_abs_diff": 5.905330181121826e-05, "name": "layer.1.k_norm", "numel": 8, "reference_max_abs": 1.0948995351791382 }, { "max_abs_diff": 9.775161743164062e-05, "mean_abs_diff": 3.227405250072479e-05, "name": "layer.1.ln1_gamma", "numel": 32, "reference_max_abs": 1.0266855955123901 }, { "max_abs_diff": 6.663799285888672e-05, "mean_abs_diff": 2.5171786546707153e-05, "name": "layer.1.ln2_gamma", "numel": 32, "reference_max_abs": 1.0616428852081299 }, { "max_abs_diff": 0.00012111663818359375, "mean_abs_diff": 5.939602851867676e-05, "name": "layer.1.q_norm", "numel": 8, "reference_max_abs": 1.0890430212020874 }, { "max_abs_diff": 0.00017654197290539742, "mean_abs_diff": 2.680368925211951e-05, "name": "layer.1.w1", "numel": 4096, "reference_max_abs": 0.10052195191383362 }, { "max_abs_diff": 0.000736873596906662, "mean_abs_diff": 2.4997923901537433e-05, "name": "layer.1.w2", "numel": 2048, "reference_max_abs": 0.09041447937488556 }, { "max_abs_diff": 0.0005706353113055229, "mean_abs_diff": 4.3157073378097266e-05, "name": "layer.1.wk", "numel": 1024, "reference_max_abs": 0.07793191075325012 }, { "max_abs_diff": 0.00010831467807292938, "mean_abs_diff": 1.86294928425923e-05, "name": "layer.1.wo", "numel": 1024, "reference_max_abs": 0.06810808926820755 }, { "max_abs_diff": 0.00021736323833465576, "mean_abs_diff": 4.158732554060407e-05, "name": "layer.1.wq", "numel": 1024, "reference_max_abs": 0.07330431044101715 }, { "max_abs_diff": 0.00010992190800607204, "mean_abs_diff": 2.2568250642507337e-05, "name": "layer.1.wv", "numel": 1024, "reference_max_abs": 0.07101407647132874 }, { "max_abs_diff": 0.00010288134217262268, "mean_abs_diff": 1.2772619811585173e-05, "name": "layer.2.b1", "numel": 128, "reference_max_abs": 0.03984907269477844 }, { "max_abs_diff": 5.8426521718502045e-06, "mean_abs_diff": 1.966911440831609e-06, "name": "layer.2.b2", "numel": 32, "reference_max_abs": 0.007834016345441341 }, { "max_abs_diff": 6.769038736820221e-05, "mean_abs_diff": 3.3319563954137266e-05, "name": "layer.2.bk", "numel": 32, "reference_max_abs": 0.019985247403383255 }, { "max_abs_diff": 7.665948942303658e-06, "mean_abs_diff": 1.8465707398718223e-06, "name": "layer.2.bo", "numel": 32, "reference_max_abs": 0.007417265325784683 }, { "max_abs_diff": 7.090624421834946e-05, "mean_abs_diff": 1.7042115359799936e-05, "name": "layer.2.bq", "numel": 32, "reference_max_abs": 0.014475008472800255 }, { "max_abs_diff": 6.322748959064484e-06, "mean_abs_diff": 2.1140040189493448e-06, "name": "layer.2.bv", "numel": 32, "reference_max_abs": 0.0063237412832677364 }, { "max_abs_diff": 6.42538070678711e-05, "mean_abs_diff": 3.224611282348633e-05, "name": "layer.2.k_norm", "numel": 8, "reference_max_abs": 1.0571837425231934 }, { "max_abs_diff": 8.440017700195312e-05, "mean_abs_diff": 2.2122636437416077e-05, "name": "layer.2.ln1_gamma", "numel": 32, "reference_max_abs": 1.010573387145996 }, { "max_abs_diff": 5.7578086853027344e-05, "mean_abs_diff": 2.179853618144989e-05, "name": "layer.2.ln2_gamma", "numel": 32, "reference_max_abs": 1.044022798538208 }, { "max_abs_diff": 6.246566772460938e-05, "mean_abs_diff": 2.9131770133972168e-05, "name": "layer.2.q_norm", "numel": 8, "reference_max_abs": 1.0667052268981934 }, { "max_abs_diff": 0.0006138840690255165, "mean_abs_diff": 2.0754481738549657e-05, "name": "layer.2.w1", "numel": 4096, "reference_max_abs": 0.09481983631849289 }, { "max_abs_diff": 0.0004909969866275787, "mean_abs_diff": 2.065423541353084e-05, "name": "layer.2.w2", "numel": 2048, "reference_max_abs": 0.07509051263332367 }, { "max_abs_diff": 0.00015350151807069778, "mean_abs_diff": 2.9006030672462657e-05, "name": "layer.2.wk", "numel": 1024, "reference_max_abs": 0.06899239122867584 }, { "max_abs_diff": 0.0006334669888019562, "mean_abs_diff": 1.3391707398113795e-05, "name": "layer.2.wo", "numel": 1024, "reference_max_abs": 0.06608258932828903 }, { "max_abs_diff": 0.00023012235760688782, "mean_abs_diff": 3.074111373280175e-05, "name": "layer.2.wq", "numel": 1024, "reference_max_abs": 0.07510703057050705 }, { "max_abs_diff": 8.446443825960159e-05, "mean_abs_diff": 1.2039243301842362e-05, "name": "layer.2.wv", "numel": 1024, "reference_max_abs": 0.06262634694576263 }, { "max_abs_diff": 0.00010608043521642685, "mean_abs_diff": 9.339089956483804e-06, "name": "layer.3.b1", "numel": 128, "reference_max_abs": 0.01722049154341221 }, { "max_abs_diff": 5.600275471806526e-06, "mean_abs_diff": 2.214220785390353e-06, "name": "layer.3.b2", "numel": 32, "reference_max_abs": 0.008650066331028938 }, { "max_abs_diff": 6.709154695272446e-05, "mean_abs_diff": 2.287591269123368e-05, "name": "layer.3.bk", "numel": 32, "reference_max_abs": 0.015245847404003143 }, { "max_abs_diff": 5.910871550440788e-06, "mean_abs_diff": 2.234112116639153e-06, "name": "layer.3.bo", "numel": 32, "reference_max_abs": 0.008161828853189945 }, { "max_abs_diff": 2.3496802896261215e-05, "mean_abs_diff": 9.216182661475614e-06, "name": "layer.3.bq", "numel": 32, "reference_max_abs": 0.017541568726301193 }, { "max_abs_diff": 1.0571209713816643e-05, "mean_abs_diff": 2.9436982913466636e-06, "name": "layer.3.bv", "numel": 32, "reference_max_abs": 0.006401557940989733 }, { "max_abs_diff": 0.00012874603271484375, "mean_abs_diff": 5.008280277252197e-05, "name": "layer.3.k_norm", "numel": 8, "reference_max_abs": 1.0353866815567017 }, { "max_abs_diff": 4.470348358154297e-05, "mean_abs_diff": 1.230090856552124e-05, "name": "layer.3.ln1_gamma", "numel": 32, "reference_max_abs": 1.0086209774017334 }, { "max_abs_diff": 4.887580871582031e-05, "mean_abs_diff": 1.8658116459846497e-05, "name": "layer.3.ln2_gamma", "numel": 32, "reference_max_abs": 1.02029287815094 }, { "max_abs_diff": 0.00011861324310302734, "mean_abs_diff": 4.4971704483032227e-05, "name": "layer.3.q_norm", "numel": 8, "reference_max_abs": 1.0334206819534302 }, { "max_abs_diff": 0.0005368366837501526, "mean_abs_diff": 1.4451478818955366e-05, "name": "layer.3.w1", "numel": 4096, "reference_max_abs": 0.08424589037895203 }, { "max_abs_diff": 0.00030784192495048046, "mean_abs_diff": 1.6030237020459026e-05, "name": "layer.3.w2", "numel": 2048, "reference_max_abs": 0.08365915715694427 }, { "max_abs_diff": 0.0001877807080745697, "mean_abs_diff": 2.4378510715905577e-05, "name": "layer.3.wk", "numel": 1024, "reference_max_abs": 0.0729791596531868 }, { "max_abs_diff": 6.274320185184479e-05, "mean_abs_diff": 8.850164704199415e-06, "name": "layer.3.wo", "numel": 1024, "reference_max_abs": 0.07539719343185425 }, { "max_abs_diff": 0.0005424029659479856, "mean_abs_diff": 2.360202415729873e-05, "name": "layer.3.wq", "numel": 1024, "reference_max_abs": 0.07918677479028702 }, { "max_abs_diff": 4.189833998680115e-05, "mean_abs_diff": 8.173008609446697e-06, "name": "layer.3.wv", "numel": 1024, "reference_max_abs": 0.07110564410686493 }, { "max_abs_diff": 0.0008211173117160797, "mean_abs_diff": 6.4955352172546554e-06, "name": "output.weight", "numel": 8192, "reference_max_abs": 0.16078321635723114 }, { "max_abs_diff": 0.0024386756122112274, "mean_abs_diff": 9.72589259617962e-06, "name": "token_emb", "numel": 8192, "reference_max_abs": 0.08853042870759964 } ] }, "first_divergence_growth": { "gradient_max_abs_diff": 0.00028845667839050293, "loss_diff": 2.9087066650390625e-05, "max_param_diff": 0.0005983412265777588, "microstep": 8, "moment_max_abs_diff": 2.8852373361587524e-05, "step": 1, "worst_param": "reported_in_final_tensor_inventory" }, "first_update": { "gradient_max_abs_diff": 0.00028845667839050293, "loss_diff": 2.9087066650390625e-05, "max_param_diff": 0.0005983412265777588, "microstep": 8, "moment_max_abs_diff": 2.8852373361587524e-05, "step": 1, "worst_param": "reported_in_final_tensor_inventory" }, "max_gradient_abs_diff": 0.0038646459579467773, "max_loss_abs_diff": 0.00047326087951660156, "max_moment_abs_diff": 0.0004012342542409897, "max_selected_logits_abs_diff": 0.003879368305206299, "max_weight_abs_diff": 0.0031391754746437073, "passed": true, "tolerances": { "gradient": 0.005, "logits": 0.05, "loss": 0.02, "moment": 0.001, "parameter": 0.005 }, "trajectory": [ { "gradient_max_abs_diff": 0.00028845667839050293, "loss_diff": 2.9087066650390625e-05, "max_param_diff": 0.0005983412265777588, "microstep": 8, "moment_max_abs_diff": 2.8852373361587524e-05, "step": 1, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0007935911417007446, "loss_diff": 9.5367431640625e-06, "max_param_diff": 0.0005987249314785004, "microstep": 16, "moment_max_abs_diff": 7.920898497104645e-05, "step": 2, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006702542304992676, "loss_diff": 0.0, "max_param_diff": 0.0005990061908960342, "microstep": 24, "moment_max_abs_diff": 0.00013794004917144775, "step": 3, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00030969083309173584, "loss_diff": 4.76837158203125e-06, "max_param_diff": 0.000599278137087822, "microstep": 32, "moment_max_abs_diff": 0.00015382468700408936, "step": 4, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005945488810539246, "loss_diff": 3.0040740966796875e-05, "max_param_diff": 0.0005993321537971497, "microstep": 40, "moment_max_abs_diff": 0.00019450485706329346, "step": 5, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00034233182668685913, "loss_diff": 6.198883056640625e-06, "max_param_diff": 0.0005992110818624496, "microstep": 48, "moment_max_abs_diff": 0.00019665062427520752, "step": 6, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00023493170738220215, "loss_diff": 2.002716064453125e-05, "max_param_diff": 0.0005991850048303604, "microstep": 56, "moment_max_abs_diff": 0.00019009411334991455, "step": 7, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00033243000507354736, "loss_diff": 9.059906005859375e-06, "max_param_diff": 0.0008239392191171646, "microstep": 64, "moment_max_abs_diff": 0.00020432472229003906, "step": 8, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00013983994722366333, "loss_diff": 9.5367431640625e-07, "max_param_diff": 0.0010368302464485168, "microstep": 72, "moment_max_abs_diff": 0.00019654631614685059, "step": 9, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00019923970103263855, "loss_diff": 1.5735626220703125e-05, "max_param_diff": 0.0012267716228961945, "microstep": 80, "moment_max_abs_diff": 0.00019310414791107178, "step": 10, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015657395124435425, "loss_diff": 2.765655517578125e-05, "max_param_diff": 0.0013969428837299347, "microstep": 88, "moment_max_abs_diff": 0.00018928200006484985, "step": 11, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016771256923675537, "loss_diff": 2.002716064453125e-05, "max_param_diff": 0.0015499088913202286, "microstep": 96, "moment_max_abs_diff": 0.00016203895211219788, "step": 12, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015157461166381836, "loss_diff": 2.384185791015625e-06, "max_param_diff": 0.0016877688467502594, "microstep": 104, "moment_max_abs_diff": 0.00014811381697654724, "step": 13, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001587122678756714, "loss_diff": 1.1444091796875e-05, "max_param_diff": 0.0018122829496860504, "microstep": 112, "moment_max_abs_diff": 0.0001246929168701172, "step": 14, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001637786626815796, "loss_diff": 4.76837158203125e-07, "max_param_diff": 0.0019249357283115387, "microstep": 120, "moment_max_abs_diff": 0.00011742860078811646, "step": 15, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016419589519500732, "loss_diff": 3.719329833984375e-05, "max_param_diff": 0.0020270030945539474, "microstep": 128, "moment_max_abs_diff": 0.00010283943265676498, "step": 16, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000147894024848938, "loss_diff": 2.193450927734375e-05, "max_param_diff": 0.002119584009051323, "microstep": 136, "moment_max_abs_diff": 9.627640247344971e-05, "step": 17, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017908215522766113, "loss_diff": 1.8596649169921875e-05, "max_param_diff": 0.0022036414593458176, "microstep": 144, "moment_max_abs_diff": 0.00010455399751663208, "step": 18, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00014056265354156494, "loss_diff": 3.337860107421875e-06, "max_param_diff": 0.002280019223690033, "microstep": 152, "moment_max_abs_diff": 0.00010815262794494629, "step": 19, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00012925267219543457, "loss_diff": 2.384185791015625e-06, "max_param_diff": 0.0023494623601436615, "microstep": 160, "moment_max_abs_diff": 0.00011026114225387573, "step": 20, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00012581050395965576, "loss_diff": 9.5367431640625e-06, "max_param_diff": 0.00241263210773468, "microstep": 168, "moment_max_abs_diff": 0.00010325759649276733, "step": 21, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002505290322005749, "loss_diff": 2.2411346435546875e-05, "max_param_diff": 0.002470117062330246, "microstep": 176, "moment_max_abs_diff": 0.00010181218385696411, "step": 22, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00012558698654174805, "loss_diff": 5.7220458984375e-06, "max_param_diff": 0.0025224480777978897, "microstep": 184, "moment_max_abs_diff": 0.00010418891906738281, "step": 23, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015157461166381836, "loss_diff": 6.67572021484375e-06, "max_param_diff": 0.002570096403360367, "microstep": 192, "moment_max_abs_diff": 0.00010892748832702637, "step": 24, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00013999082148075104, "loss_diff": 4.291534423828125e-06, "max_param_diff": 0.0026134904474020004, "microstep": 200, "moment_max_abs_diff": 0.00011041015386581421, "step": 25, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015145912766456604, "loss_diff": 8.58306884765625e-06, "max_param_diff": 0.002653013914823532, "microstep": 208, "moment_max_abs_diff": 0.0001134723424911499, "step": 26, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017678644508123398, "loss_diff": 7.62939453125e-06, "max_param_diff": 0.0026890169829130173, "microstep": 216, "moment_max_abs_diff": 0.00011461973190307617, "step": 27, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001447596587240696, "loss_diff": 1.4781951904296875e-05, "max_param_diff": 0.0027218125760555267, "microstep": 224, "moment_max_abs_diff": 0.00011357665061950684, "step": 28, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016600452363491058, "loss_diff": 7.152557373046875e-06, "max_param_diff": 0.002751687541604042, "microstep": 232, "moment_max_abs_diff": 0.0001128464937210083, "step": 29, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020315777510404587, "loss_diff": 3.337860107421875e-06, "max_param_diff": 0.0027789007872343063, "microstep": 240, "moment_max_abs_diff": 0.00011477794032543898, "step": 30, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015337206423282623, "loss_diff": 1.430511474609375e-05, "max_param_diff": 0.002803688868880272, "microstep": 248, "moment_max_abs_diff": 0.00011478806845843792, "step": 31, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00012380816042423248, "loss_diff": 1.1920928955078125e-05, "max_param_diff": 0.0028262678533792496, "microstep": 256, "moment_max_abs_diff": 0.00011569005437195301, "step": 32, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001656584208831191, "loss_diff": 2.4318695068359375e-05, "max_param_diff": 0.0028468314558267593, "microstep": 264, "moment_max_abs_diff": 0.0001206868328154087, "step": 33, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018573366105556488, "loss_diff": 1.4781951904296875e-05, "max_param_diff": 0.0028655584901571274, "microstep": 272, "moment_max_abs_diff": 0.00012719165533781052, "step": 34, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017773604486137629, "loss_diff": 1.1920928955078125e-05, "max_param_diff": 0.002882611006498337, "microstep": 280, "moment_max_abs_diff": 0.00013224594295024872, "step": 35, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018497928977012634, "loss_diff": 7.62939453125e-06, "max_param_diff": 0.002898136153817177, "microstep": 288, "moment_max_abs_diff": 0.0001375193241983652, "step": 36, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00022127665579319, "loss_diff": 3.337860107421875e-06, "max_param_diff": 0.002912268042564392, "microstep": 296, "moment_max_abs_diff": 0.00014143926091492176, "step": 37, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00021388661116361618, "loss_diff": 1.3828277587890625e-05, "max_param_diff": 0.0029251333326101303, "microstep": 304, "moment_max_abs_diff": 0.00014868401922285557, "step": 38, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002012699842453003, "loss_diff": 1.52587890625e-05, "max_param_diff": 0.002936840057373047, "microstep": 312, "moment_max_abs_diff": 0.00014549982734024525, "step": 39, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020934641361236572, "loss_diff": 1.049041748046875e-05, "max_param_diff": 0.002947494387626648, "microstep": 320, "moment_max_abs_diff": 0.000141938216984272, "step": 40, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015074270777404308, "loss_diff": 1.430511474609375e-05, "max_param_diff": 0.002957187592983246, "microstep": 328, "moment_max_abs_diff": 0.00013684015721082687, "step": 41, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00014242716133594513, "loss_diff": 9.059906005859375e-06, "max_param_diff": 0.002966005355119705, "microstep": 336, "moment_max_abs_diff": 0.00013687554746866226, "step": 42, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00025212671607732773, "loss_diff": 2.0503997802734375e-05, "max_param_diff": 0.0029740240424871445, "microstep": 344, "moment_max_abs_diff": 0.00014154892414808273, "step": 43, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018719956278800964, "loss_diff": 1.52587890625e-05, "max_param_diff": 0.002981318160891533, "microstep": 352, "moment_max_abs_diff": 0.0001461142674088478, "step": 44, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000291372649371624, "loss_diff": 2.6226043701171875e-05, "max_param_diff": 0.0029814522713422775, "microstep": 360, "moment_max_abs_diff": 0.00015045609325170517, "step": 45, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00027837371453642845, "loss_diff": 1.621246337890625e-05, "max_param_diff": 0.002982471138238907, "microstep": 368, "moment_max_abs_diff": 0.00015915697440505028, "step": 46, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001987135037779808, "loss_diff": 9.059906005859375e-06, "max_param_diff": 0.003000974655151367, "microstep": 376, "moment_max_abs_diff": 0.00016019120812416077, "step": 47, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020660972222685814, "loss_diff": 2.7179718017578125e-05, "max_param_diff": 0.003017786890268326, "microstep": 384, "moment_max_abs_diff": 0.00016326364129781723, "step": 48, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001879492774605751, "loss_diff": 7.152557373046875e-06, "max_param_diff": 0.0030330605804920197, "microstep": 392, "moment_max_abs_diff": 0.00015453435480594635, "step": 49, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003290558233857155, "loss_diff": 4.76837158203125e-07, "max_param_diff": 0.003046933561563492, "microstep": 400, "moment_max_abs_diff": 0.0001666559837758541, "step": 50, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002720039337873459, "loss_diff": 2.9087066650390625e-05, "max_param_diff": 0.00305953249335289, "microstep": 408, "moment_max_abs_diff": 0.0001771911047399044, "step": 51, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002995189279317856, "loss_diff": 3.8623809814453125e-05, "max_param_diff": 0.003070969134569168, "microstep": 416, "moment_max_abs_diff": 0.00018942402675747871, "step": 52, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00032617710530757904, "loss_diff": 8.106231689453125e-06, "max_param_diff": 0.00308135524392128, "microstep": 424, "moment_max_abs_diff": 0.0002030995674431324, "step": 53, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00033145956695079803, "loss_diff": 1.621246337890625e-05, "max_param_diff": 0.0030907802283763885, "microstep": 432, "moment_max_abs_diff": 0.00021593552082777023, "step": 54, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003264574334025383, "loss_diff": 5.245208740234375e-06, "max_param_diff": 0.003099333494901657, "microstep": 440, "moment_max_abs_diff": 0.00022698752582073212, "step": 55, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003517698496580124, "loss_diff": 8.106231689453125e-06, "max_param_diff": 0.0031070969998836517, "microstep": 448, "moment_max_abs_diff": 0.0002394663169980049, "step": 56, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00039798393845558167, "loss_diff": 2.7179718017578125e-05, "max_param_diff": 0.0031070969998836517, "microstep": 456, "moment_max_abs_diff": 0.000255318358540535, "step": 57, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00034826435148715973, "loss_diff": 4.76837158203125e-07, "max_param_diff": 0.0031070932745933533, "microstep": 464, "moment_max_abs_diff": 0.00026461388915777206, "step": 58, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003654472529888153, "loss_diff": 4.1484832763671875e-05, "max_param_diff": 0.003107089549303055, "microstep": 472, "moment_max_abs_diff": 0.00027469824999570847, "step": 59, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003785751760005951, "loss_diff": 1.430511474609375e-06, "max_param_diff": 0.0031070858240127563, "microstep": 480, "moment_max_abs_diff": 0.00026474427431821823, "step": 60, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00029511749744415283, "loss_diff": 8.58306884765625e-06, "max_param_diff": 0.0031070858240127563, "microstep": 488, "moment_max_abs_diff": 0.0002665901556611061, "step": 61, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000385187566280365, "loss_diff": 6.67572021484375e-06, "max_param_diff": 0.0031070858240127563, "microstep": 496, "moment_max_abs_diff": 0.000278448686003685, "step": 62, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00046667829155921936, "loss_diff": 3.6716461181640625e-05, "max_param_diff": 0.003107082098722458, "microstep": 504, "moment_max_abs_diff": 0.00029727257788181305, "step": 63, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00042825937271118164, "loss_diff": 1.239776611328125e-05, "max_param_diff": 0.0031070783734321594, "microstep": 512, "moment_max_abs_diff": 0.00031037256121635437, "step": 64, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000492926687002182, "loss_diff": 2.8133392333984375e-05, "max_param_diff": 0.003107074648141861, "microstep": 520, "moment_max_abs_diff": 0.000328628346323967, "step": 65, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005018003284931183, "loss_diff": 2.86102294921875e-06, "max_param_diff": 0.003107067197561264, "microstep": 528, "moment_max_abs_diff": 0.0003459472209215164, "step": 66, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00038478150963783264, "loss_diff": 2.4318695068359375e-05, "max_param_diff": 0.0031070634722709656, "microstep": 536, "moment_max_abs_diff": 0.00034958869218826294, "step": 67, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004339143633842468, "loss_diff": 5.340576171875e-05, "max_param_diff": 0.0031070560216903687, "microstep": 544, "moment_max_abs_diff": 0.00035732798278331757, "step": 68, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004911571741104126, "loss_diff": 9.5367431640625e-06, "max_param_diff": 0.00310705229640007, "microstep": 552, "moment_max_abs_diff": 0.000370711088180542, "step": 69, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003534369170665741, "loss_diff": 4.5299530029296875e-05, "max_param_diff": 0.0031070485711097717, "microstep": 560, "moment_max_abs_diff": 0.00036652758717536926, "step": 70, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00035600364208221436, "loss_diff": 3.5762786865234375e-05, "max_param_diff": 0.003107041120529175, "microstep": 568, "moment_max_abs_diff": 0.0003638267517089844, "step": 71, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00035610422492027283, "loss_diff": 3.9577484130859375e-05, "max_param_diff": 0.003107033669948578, "microstep": 576, "moment_max_abs_diff": 0.00036305561661720276, "step": 72, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00042630359530448914, "loss_diff": 4.5299530029296875e-05, "max_param_diff": 0.003107026219367981, "microstep": 584, "moment_max_abs_diff": 0.0003693811595439911, "step": 73, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00039043650031089783, "loss_diff": 3.0517578125e-05, "max_param_diff": 0.003107018768787384, "microstep": 592, "moment_max_abs_diff": 0.00037148594856262207, "step": 74, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003841817378997803, "loss_diff": 5.817413330078125e-05, "max_param_diff": 0.003107011318206787, "microstep": 600, "moment_max_abs_diff": 0.00037275440990924835, "step": 75, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004429146647453308, "loss_diff": 6.008148193359375e-05, "max_param_diff": 0.00310700386762619, "microstep": 608, "moment_max_abs_diff": 0.00037277303636074066, "step": 76, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00043088942766189575, "loss_diff": 6.389617919921875e-05, "max_param_diff": 0.0031069964170455933, "microstep": 616, "moment_max_abs_diff": 0.00037858448922634125, "step": 77, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000371783971786499, "loss_diff": 5.245208740234375e-06, "max_param_diff": 0.003106992691755295, "microstep": 624, "moment_max_abs_diff": 0.0003667417913675308, "step": 78, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000339265912771225, "loss_diff": 3.2901763916015625e-05, "max_param_diff": 0.003106985241174698, "microstep": 632, "moment_max_abs_diff": 0.00033373385667800903, "step": 79, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00025022029876708984, "loss_diff": 1.811981201171875e-05, "max_param_diff": 0.003106977790594101, "microstep": 640, "moment_max_abs_diff": 0.00031045079231262207, "step": 80, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00023104995489120483, "loss_diff": 6.67572021484375e-06, "max_param_diff": 0.0031069740653038025, "microstep": 648, "moment_max_abs_diff": 0.0002976655960083008, "step": 81, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00046940892934799194, "loss_diff": 3.6716461181640625e-05, "max_param_diff": 0.0031069666147232056, "microstep": 656, "moment_max_abs_diff": 0.0002970695495605469, "step": 82, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00031023845076560974, "loss_diff": 2.4318695068359375e-05, "max_param_diff": 0.0031069591641426086, "microstep": 664, "moment_max_abs_diff": 0.0002983883023262024, "step": 83, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000400688499212265, "loss_diff": 6.151199340820312e-05, "max_param_diff": 0.0031069479882717133, "microstep": 672, "moment_max_abs_diff": 0.00029885396361351013, "step": 84, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004471726715564728, "loss_diff": 2.956390380859375e-05, "max_param_diff": 0.0031069405376911163, "microstep": 680, "moment_max_abs_diff": 0.00031047873198986053, "step": 85, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003222525119781494, "loss_diff": 3.7670135498046875e-05, "max_param_diff": 0.003106929361820221, "microstep": 688, "moment_max_abs_diff": 0.0003004297614097595, "step": 86, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002853795886039734, "loss_diff": 4.76837158203125e-05, "max_param_diff": 0.003106921911239624, "microstep": 696, "moment_max_abs_diff": 0.000295441597700119, "step": 87, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00023928657174110413, "loss_diff": 2.9087066650390625e-05, "max_param_diff": 0.0031069107353687286, "microstep": 704, "moment_max_abs_diff": 0.0002777688205242157, "step": 88, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004751235246658325, "loss_diff": 2.2411346435546875e-05, "max_param_diff": 0.0031068995594978333, "microstep": 712, "moment_max_abs_diff": 0.0002919342368841171, "step": 89, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00035123899579048157, "loss_diff": 7.200241088867188e-05, "max_param_diff": 0.003106888383626938, "microstep": 720, "moment_max_abs_diff": 0.0002978648990392685, "step": 90, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00042729079723358154, "loss_diff": 6.389617919921875e-05, "max_param_diff": 0.0031068772077560425, "microstep": 728, "moment_max_abs_diff": 0.0003108084201812744, "step": 91, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00045010074973106384, "loss_diff": 6.723403930664062e-05, "max_param_diff": 0.003106866031885147, "microstep": 736, "moment_max_abs_diff": 0.0003247372806072235, "step": 92, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00040648505091667175, "loss_diff": 3.0040740966796875e-05, "max_param_diff": 0.00310685858130455, "microstep": 744, "moment_max_abs_diff": 0.00033291056752204895, "step": 93, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00043214112520217896, "loss_diff": 2.002716064453125e-05, "max_param_diff": 0.003106847405433655, "microstep": 752, "moment_max_abs_diff": 0.0003428347408771515, "step": 94, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004316754639148712, "loss_diff": 7.43865966796875e-05, "max_param_diff": 0.003106839954853058, "microstep": 760, "moment_max_abs_diff": 0.00035171955823898315, "step": 95, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004985928535461426, "loss_diff": 4.863739013671875e-05, "max_param_diff": 0.0030911494977772236, "microstep": 768, "moment_max_abs_diff": 0.00036640465259552, "step": 96, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00042780861258506775, "loss_diff": 3.24249267578125e-05, "max_param_diff": 0.0030939863063395023, "microstep": 776, "moment_max_abs_diff": 0.00037254393100738525, "step": 97, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004056990146636963, "loss_diff": 3.24249267578125e-05, "max_param_diff": 0.003096552100032568, "microstep": 784, "moment_max_abs_diff": 0.00037585943937301636, "step": 98, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004060491919517517, "loss_diff": 2.384185791015625e-05, "max_param_diff": 0.003098873421549797, "microstep": 792, "moment_max_abs_diff": 0.0003623552620410919, "step": 99, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003277920186519623, "loss_diff": 8.106231689453125e-06, "max_param_diff": 0.0031009730882942677, "microstep": 800, "moment_max_abs_diff": 0.000349290668964386, "step": 100, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00042128562927246094, "loss_diff": 2.86102294921875e-05, "max_param_diff": 0.0031028715893626213, "microstep": 808, "moment_max_abs_diff": 0.00035649165511131287, "step": 101, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00048479437828063965, "loss_diff": 5.14984130859375e-05, "max_param_diff": 0.003104587085545063, "microstep": 816, "moment_max_abs_diff": 0.0003693215548992157, "step": 102, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000420592725276947, "loss_diff": 4.863739013671875e-05, "max_param_diff": 0.0031061386689543724, "microstep": 824, "moment_max_abs_diff": 0.00037445127964019775, "step": 103, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005120411515235901, "loss_diff": 7.033348083496094e-05, "max_param_diff": 0.003107542172074318, "microstep": 832, "moment_max_abs_diff": 0.00038821250200271606, "step": 104, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004876852035522461, "loss_diff": 6.008148193359375e-05, "max_param_diff": 0.0031088097020983696, "microstep": 840, "moment_max_abs_diff": 0.00039816275238990784, "step": 105, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003835931420326233, "loss_diff": 4.1961669921875e-05, "max_param_diff": 0.0031099561601877213, "microstep": 848, "moment_max_abs_diff": 0.00039670616388320923, "step": 106, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003899410367012024, "loss_diff": 5.435943603515625e-05, "max_param_diff": 0.0031109917908906937, "microstep": 856, "moment_max_abs_diff": 0.00039358437061309814, "step": 107, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00044633448123931885, "loss_diff": 5.9604644775390625e-05, "max_param_diff": 0.0031119268387556076, "microstep": 864, "moment_max_abs_diff": 0.0003988593816757202, "step": 108, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00033463165163993835, "loss_diff": 4.267692565917969e-05, "max_param_diff": 0.0031127724796533585, "microstep": 872, "moment_max_abs_diff": 0.00038838014006614685, "step": 109, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00034111738204956055, "loss_diff": 5.245208740234375e-05, "max_param_diff": 0.003113536164164543, "microstep": 880, "moment_max_abs_diff": 0.0003836527466773987, "step": 110, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002486035227775574, "loss_diff": 3.5762786865234375e-05, "max_param_diff": 0.0031142262741923332, "microstep": 888, "moment_max_abs_diff": 0.00037014856934547424, "step": 111, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003808438777923584, "loss_diff": 5.793571472167969e-05, "max_param_diff": 0.003114849328994751, "microstep": 896, "moment_max_abs_diff": 0.0003712177276611328, "step": 112, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000351686030626297, "loss_diff": 5.459785461425781e-05, "max_param_diff": 0.003115410916507244, "microstep": 904, "moment_max_abs_diff": 0.00036926567554473877, "step": 113, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003103017807006836, "loss_diff": 5.054473876953125e-05, "max_param_diff": 0.0031159184873104095, "microstep": 912, "moment_max_abs_diff": 0.00036336854100227356, "step": 114, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004308074712753296, "loss_diff": 6.318092346191406e-05, "max_param_diff": 0.0031163757666945457, "microstep": 920, "moment_max_abs_diff": 0.0003618299961090088, "step": 115, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00033789128065109253, "loss_diff": 4.9114227294921875e-05, "max_param_diff": 0.0031167883425951004, "microstep": 928, "moment_max_abs_diff": 0.0003594346344470978, "step": 116, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00029521435499191284, "loss_diff": 2.47955322265625e-05, "max_param_diff": 0.003117159940302372, "microstep": 936, "moment_max_abs_diff": 0.00034978240728378296, "step": 117, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00030966848134994507, "loss_diff": 5.245208740234375e-06, "max_param_diff": 0.003117496147751808, "microstep": 944, "moment_max_abs_diff": 0.000315200537443161, "step": 118, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00019517168402671814, "loss_diff": 1.1444091796875e-05, "max_param_diff": 0.003117798827588558, "microstep": 952, "moment_max_abs_diff": 0.0002772994339466095, "step": 119, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020256265997886658, "loss_diff": 2.09808349609375e-05, "max_param_diff": 0.0031180717051029205, "microstep": 960, "moment_max_abs_diff": 0.0002574734389781952, "step": 120, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00034189969301223755, "loss_diff": 3.552436828613281e-05, "max_param_diff": 0.003118317574262619, "microstep": 968, "moment_max_abs_diff": 0.0002521537244319916, "step": 121, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00024650245904922485, "loss_diff": 4.673004150390625e-05, "max_param_diff": 0.0031185392290353775, "microstep": 976, "moment_max_abs_diff": 0.00025159120559692383, "step": 122, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002713054418563843, "loss_diff": 3.647804260253906e-05, "max_param_diff": 0.003118738532066345, "microstep": 984, "moment_max_abs_diff": 0.0002535618841648102, "step": 123, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00034867972135543823, "loss_diff": 1.811981201171875e-05, "max_param_diff": 0.0031189173460006714, "microstep": 992, "moment_max_abs_diff": 0.00026307255029678345, "step": 124, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002415291965007782, "loss_diff": 4.100799560546875e-05, "max_param_diff": 0.0031190775334835052, "microstep": 1000, "moment_max_abs_diff": 0.00025168806314468384, "step": 125, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020229443907737732, "loss_diff": 3.719329833984375e-05, "max_param_diff": 0.0031192218884825706, "microstep": 1008, "moment_max_abs_diff": 0.00024391710758209229, "step": 126, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002537071704864502, "loss_diff": 1.0013580322265625e-05, "max_param_diff": 0.003119351342320442, "microstep": 1016, "moment_max_abs_diff": 0.00022182241082191467, "step": 127, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00035656988620758057, "loss_diff": 2.09808349609375e-05, "max_param_diff": 0.0031194668263196945, "microstep": 1024, "moment_max_abs_diff": 0.00023529678583145142, "step": 128, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002949051558971405, "loss_diff": 6.127357482910156e-05, "max_param_diff": 0.003119570203125477, "microstep": 1032, "moment_max_abs_diff": 0.00024125725030899048, "step": 129, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003287121653556824, "loss_diff": 2.765655517578125e-05, "max_param_diff": 0.0031196633353829384, "microstep": 1040, "moment_max_abs_diff": 0.00025000423192977905, "step": 130, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003399774432182312, "loss_diff": 4.601478576660156e-05, "max_param_diff": 0.0031197471544146538, "microstep": 1048, "moment_max_abs_diff": 0.00025900453329086304, "step": 131, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00030077993869781494, "loss_diff": 3.8623809814453125e-05, "max_param_diff": 0.003119821660220623, "microstep": 1056, "moment_max_abs_diff": 0.00026318058371543884, "step": 132, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00033280253410339355, "loss_diff": 3.600120544433594e-05, "max_param_diff": 0.0031198877841234207, "microstep": 1064, "moment_max_abs_diff": 0.00027014315128326416, "step": 133, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002924874424934387, "loss_diff": 4.363059997558594e-05, "max_param_diff": 0.0031199464574456215, "microstep": 1072, "moment_max_abs_diff": 0.0002723783254623413, "step": 134, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00036606937646865845, "loss_diff": 1.7642974853515625e-05, "max_param_diff": 0.0031210314482450485, "microstep": 1080, "moment_max_abs_diff": 0.000281747430562973, "step": 135, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003069266676902771, "loss_diff": 4.7206878662109375e-05, "max_param_diff": 0.003122011199593544, "microstep": 1088, "moment_max_abs_diff": 0.0002842657268047333, "step": 136, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002787858247756958, "loss_diff": 1.9550323486328125e-05, "max_param_diff": 0.0031228959560394287, "microstep": 1096, "moment_max_abs_diff": 0.0002837181091308594, "step": 137, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002512335777282715, "loss_diff": 1.2159347534179688e-05, "max_param_diff": 0.003123694099485874, "microstep": 1104, "moment_max_abs_diff": 0.0002770274877548218, "step": 138, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020035728812217712, "loss_diff": 1.5735626220703125e-05, "max_param_diff": 0.003124414011836052, "microstep": 1112, "moment_max_abs_diff": 0.00026065483689308167, "step": 139, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00028581172227859497, "loss_diff": 3.4332275390625e-05, "max_param_diff": 0.0031250640749931335, "microstep": 1120, "moment_max_abs_diff": 0.0002631731331348419, "step": 140, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00036407262086868286, "loss_diff": 5.030632019042969e-05, "max_param_diff": 0.0031256498768925667, "microstep": 1128, "moment_max_abs_diff": 0.00027326494455337524, "step": 141, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002739429473876953, "loss_diff": 3.981590270996094e-05, "max_param_diff": 0.0031261788681149483, "microstep": 1136, "moment_max_abs_diff": 0.00027333199977874756, "step": 142, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003549158573150635, "loss_diff": 3.4809112548828125e-05, "max_param_diff": 0.0031266557052731514, "microstep": 1144, "moment_max_abs_diff": 0.0002814978361129761, "step": 143, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003097057342529297, "loss_diff": 2.9087066650390625e-05, "max_param_diff": 0.003127085044980049, "microstep": 1152, "moment_max_abs_diff": 0.0002843216061592102, "step": 144, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000266052782535553, "loss_diff": 2.1219253540039062e-05, "max_param_diff": 0.0031274715438485146, "microstep": 1160, "moment_max_abs_diff": 0.00028249621391296387, "step": 145, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002727210521697998, "loss_diff": 3.0279159545898438e-05, "max_param_diff": 0.0031278198584914207, "microstep": 1168, "moment_max_abs_diff": 0.00028152018785476685, "step": 146, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002860724925994873, "loss_diff": 2.9325485229492188e-05, "max_param_diff": 0.003128133714199066, "microstep": 1176, "moment_max_abs_diff": 0.0002819746732711792, "step": 147, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002405792474746704, "loss_diff": 3.337860107421875e-05, "max_param_diff": 0.0031284168362617493, "microstep": 1184, "moment_max_abs_diff": 0.0002735704183578491, "step": 148, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002515539526939392, "loss_diff": 3.552436828613281e-05, "max_param_diff": 0.003128672018647194, "microstep": 1192, "moment_max_abs_diff": 0.00027136504650115967, "step": 149, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001318603754043579, "loss_diff": 2.3603439331054688e-05, "max_param_diff": 0.0031289011240005493, "microstep": 1200, "moment_max_abs_diff": 0.00025741755962371826, "step": 150, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00026810914278030396, "loss_diff": 1.9311904907226562e-05, "max_param_diff": 0.0031291060149669647, "microstep": 1208, "moment_max_abs_diff": 0.00025848299264907837, "step": 151, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00022852420806884766, "loss_diff": 2.8848648071289062e-05, "max_param_diff": 0.0031292904168367386, "microstep": 1216, "moment_max_abs_diff": 0.000255487859249115, "step": 152, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00021776556968688965, "loss_diff": 2.6226043701171875e-05, "max_param_diff": 0.00312945619225502, "microstep": 1224, "moment_max_abs_diff": 0.0002517104148864746, "step": 153, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002515837550163269, "loss_diff": 2.9087066650390625e-05, "max_param_diff": 0.0031296061351895332, "microstep": 1232, "moment_max_abs_diff": 0.0002500265836715698, "step": 154, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00022573024034500122, "loss_diff": 2.5987625122070312e-05, "max_param_diff": 0.003129740245640278, "microstep": 1240, "moment_max_abs_diff": 0.00024759769439697266, "step": 155, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020381808280944824, "loss_diff": 9.5367431640625e-06, "max_param_diff": 0.0031298603862524033, "microstep": 1248, "moment_max_abs_diff": 0.00024322420358657837, "step": 156, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00019902735948562622, "loss_diff": 1.1920928955078125e-06, "max_param_diff": 0.003129967488348484, "microstep": 1256, "moment_max_abs_diff": 0.00021686404943466187, "step": 157, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00013985484838485718, "loss_diff": 3.814697265625e-06, "max_param_diff": 0.0031300634145736694, "microstep": 1264, "moment_max_abs_diff": 0.00018836185336112976, "step": 158, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00014930963516235352, "loss_diff": 2.4318695068359375e-05, "max_param_diff": 0.003130149096250534, "microstep": 1272, "moment_max_abs_diff": 0.00016330182552337646, "step": 159, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018835067749023438, "loss_diff": 1.0967254638671875e-05, "max_param_diff": 0.0031302254647016525, "microstep": 1280, "moment_max_abs_diff": 0.00016472116112709045, "step": 160, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001771971583366394, "loss_diff": 3.3855438232421875e-05, "max_param_diff": 0.0031302934512495995, "microstep": 1288, "moment_max_abs_diff": 0.0001659691333770752, "step": 161, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017059966921806335, "loss_diff": 4.291534423828125e-06, "max_param_diff": 0.003130353055894375, "microstep": 1296, "moment_max_abs_diff": 0.00016427040100097656, "step": 162, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000258423388004303, "loss_diff": 2.86102294921875e-06, "max_param_diff": 0.003130406141281128, "microstep": 1304, "moment_max_abs_diff": 0.00017368420958518982, "step": 163, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001662522554397583, "loss_diff": 4.291534423828125e-06, "max_param_diff": 0.0031304536387324333, "microstep": 1312, "moment_max_abs_diff": 0.00016625598073005676, "step": 164, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00014572963118553162, "loss_diff": 1.9073486328125e-06, "max_param_diff": 0.003130495548248291, "microstep": 1320, "moment_max_abs_diff": 0.00016420334577560425, "step": 165, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002314746379852295, "loss_diff": 2.8133392333984375e-05, "max_param_diff": 0.003130531869828701, "microstep": 1328, "moment_max_abs_diff": 0.00014654546976089478, "step": 166, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002387426793575287, "loss_diff": 1.6689300537109375e-06, "max_param_diff": 0.003130565397441387, "microstep": 1336, "moment_max_abs_diff": 0.000155765563249588, "step": 167, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00021531060338020325, "loss_diff": 2.0265579223632812e-05, "max_param_diff": 0.0031305942684412003, "microstep": 1344, "moment_max_abs_diff": 0.0001617223024368286, "step": 168, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002343505620956421, "loss_diff": 2.0503997802734375e-05, "max_param_diff": 0.003130619414150715, "microstep": 1352, "moment_max_abs_diff": 0.00016898661851882935, "step": 169, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00023583322763442993, "loss_diff": 2.384185791015625e-05, "max_param_diff": 0.0031306417658925056, "microstep": 1360, "moment_max_abs_diff": 0.00017567351460456848, "step": 170, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018580257892608643, "loss_diff": 1.621246337890625e-05, "max_param_diff": 0.003130660392343998, "microstep": 1368, "moment_max_abs_diff": 0.00017668679356575012, "step": 171, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00022374093532562256, "loss_diff": 1.430511474609375e-05, "max_param_diff": 0.0031306762248277664, "microstep": 1376, "moment_max_abs_diff": 0.00018139183521270752, "step": 172, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001945197582244873, "loss_diff": 5.4836273193359375e-06, "max_param_diff": 0.0031306901946663857, "microstep": 1384, "moment_max_abs_diff": 0.00018270686268806458, "step": 173, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020869076251983643, "loss_diff": 3.814697265625e-06, "max_param_diff": 0.003131016157567501, "microstep": 1392, "moment_max_abs_diff": 0.000185307115316391, "step": 174, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020462274551391602, "loss_diff": 7.3909759521484375e-06, "max_param_diff": 0.003131309524178505, "microstep": 1400, "moment_max_abs_diff": 0.00018723681569099426, "step": 175, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018345564603805542, "loss_diff": 1.239776611328125e-05, "max_param_diff": 0.003131573088467121, "microstep": 1408, "moment_max_abs_diff": 0.00018685311079025269, "step": 176, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001609399914741516, "loss_diff": 1.1205673217773438e-05, "max_param_diff": 0.0031318115070462227, "microstep": 1416, "moment_max_abs_diff": 0.0001842603087425232, "step": 177, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00012981146574020386, "loss_diff": 2.1457672119140625e-06, "max_param_diff": 0.0031320247799158096, "microstep": 1424, "moment_max_abs_diff": 0.00017038360238075256, "step": 178, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020479410886764526, "loss_diff": 2.1457672119140625e-06, "max_param_diff": 0.003132217563688755, "microstep": 1432, "moment_max_abs_diff": 0.0001738220453262329, "step": 179, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00024124979972839355, "loss_diff": 6.198883056640625e-06, "max_param_diff": 0.003132389858365059, "microstep": 1440, "moment_max_abs_diff": 0.0001805722713470459, "step": 180, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018002837896347046, "loss_diff": 4.76837158203125e-07, "max_param_diff": 0.003132544457912445, "microstep": 1448, "moment_max_abs_diff": 0.00018052011728286743, "step": 181, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00022440403699874878, "loss_diff": 8.106231689453125e-06, "max_param_diff": 0.0031326832249760628, "microstep": 1456, "moment_max_abs_diff": 0.00018490850925445557, "step": 182, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017604976892471313, "loss_diff": 1.430511474609375e-06, "max_param_diff": 0.0031328080222010612, "microstep": 1464, "moment_max_abs_diff": 0.00018402934074401855, "step": 183, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016769766807556152, "loss_diff": 8.106231689453125e-06, "max_param_diff": 0.003132917918264866, "microstep": 1472, "moment_max_abs_diff": 0.00018239766359329224, "step": 184, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018680840730667114, "loss_diff": 4.76837158203125e-06, "max_param_diff": 0.0031330185011029243, "microstep": 1480, "moment_max_abs_diff": 0.00018283724784851074, "step": 185, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00019288063049316406, "loss_diff": 1.4543533325195312e-05, "max_param_diff": 0.003133106976747513, "microstep": 1488, "moment_max_abs_diff": 0.00018384307622909546, "step": 186, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001447722315788269, "loss_diff": 5.0067901611328125e-06, "max_param_diff": 0.003133186139166355, "microstep": 1496, "moment_max_abs_diff": 0.00017958134412765503, "step": 187, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015825778245925903, "loss_diff": 4.0531158447265625e-06, "max_param_diff": 0.0031332559883594513, "microstep": 1504, "moment_max_abs_diff": 0.00017745047807693481, "step": 188, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00012321025133132935, "loss_diff": 1.621246337890625e-05, "max_param_diff": 0.0031333183869719505, "microstep": 1512, "moment_max_abs_diff": 0.00016775727272033691, "step": 189, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020033866167068481, "loss_diff": 2.1457672119140625e-06, "max_param_diff": 0.0031333742663264275, "microstep": 1520, "moment_max_abs_diff": 0.00017102062702178955, "step": 190, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015931576490402222, "loss_diff": 9.775161743164062e-06, "max_param_diff": 0.0031334245577454567, "microstep": 1528, "moment_max_abs_diff": 0.00016911327838897705, "step": 191, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001737065613269806, "loss_diff": 2.384185791015625e-06, "max_param_diff": 0.003133467398583889, "microstep": 1536, "moment_max_abs_diff": 0.00016957521438598633, "step": 192, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016070902347564697, "loss_diff": 3.337860107421875e-06, "max_param_diff": 0.003133506514132023, "microstep": 1544, "moment_max_abs_diff": 0.0001682564616203308, "step": 193, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016796961426734924, "loss_diff": 1.2159347534179688e-05, "max_param_diff": 0.003133540041744709, "microstep": 1552, "moment_max_abs_diff": 0.00016823410987854004, "step": 194, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00012174993753433228, "loss_diff": 9.059906005859375e-06, "max_param_diff": 0.0031335698440670967, "microstep": 1560, "moment_max_abs_diff": 0.00016358494758605957, "step": 195, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00013032183051109314, "loss_diff": 1.7642974853515625e-05, "max_param_diff": 0.0031335949897766113, "microstep": 1568, "moment_max_abs_diff": 0.00014823675155639648, "step": 196, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016846880316734314, "loss_diff": 1.5974044799804688e-05, "max_param_diff": 0.003133617341518402, "microstep": 1576, "moment_max_abs_diff": 0.0001271963119506836, "step": 197, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00010695122182369232, "loss_diff": 1.239776611328125e-05, "max_param_diff": 0.0031336378306150436, "microstep": 1584, "moment_max_abs_diff": 0.0001079849898815155, "step": 198, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017428025603294373, "loss_diff": 3.814697265625e-06, "max_param_diff": 0.0031336555257439613, "microstep": 1592, "moment_max_abs_diff": 0.00010849162936210632, "step": 199, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00011835992336273193, "loss_diff": 5.0067901611328125e-06, "max_param_diff": 0.0031336694955825806, "microstep": 1600, "moment_max_abs_diff": 0.00010947883129119873, "step": 200, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00012180954217910767, "loss_diff": 2.4080276489257812e-05, "max_param_diff": 0.0031336816027760506, "microstep": 1608, "moment_max_abs_diff": 0.00011071190237998962, "step": 201, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002351701259613037, "loss_diff": 8.344650268554688e-06, "max_param_diff": 0.0031336909160017967, "microstep": 1616, "moment_max_abs_diff": 0.00012315809726715088, "step": 202, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00021591410040855408, "loss_diff": 2.3365020751953125e-05, "max_param_diff": 0.0031336992979049683, "microstep": 1624, "moment_max_abs_diff": 0.0001214742660522461, "step": 203, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001733154058456421, "loss_diff": 3.0517578125e-05, "max_param_diff": 0.0031337058171629906, "microstep": 1632, "moment_max_abs_diff": 0.00012022629380226135, "step": 204, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00024481117725372314, "loss_diff": 2.9087066650390625e-05, "max_param_diff": 0.0031337104737758636, "microstep": 1640, "moment_max_abs_diff": 0.00010607019066810608, "step": 205, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017793849110603333, "loss_diff": 9.5367431640625e-06, "max_param_diff": 0.003133714199066162, "microstep": 1648, "moment_max_abs_diff": 0.00011299923062324524, "step": 206, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00019568204879760742, "loss_diff": 2.1457672119140625e-06, "max_param_diff": 0.0031337160617113113, "microstep": 1656, "moment_max_abs_diff": 0.00011657550930976868, "step": 207, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001910477876663208, "loss_diff": 3.0994415283203125e-05, "max_param_diff": 0.0031337160617113113, "microstep": 1664, "moment_max_abs_diff": 0.00012402236461639404, "step": 208, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001452406868338585, "loss_diff": 2.0503997802734375e-05, "max_param_diff": 0.0031337160617113113, "microstep": 1672, "moment_max_abs_diff": 0.00012558698654174805, "step": 209, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016288179904222488, "loss_diff": 1.049041748046875e-05, "max_param_diff": 0.003133714199066162, "microstep": 1680, "moment_max_abs_diff": 0.00012372061610221863, "step": 210, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016049295663833618, "loss_diff": 1.049041748046875e-05, "max_param_diff": 0.003133712336421013, "microstep": 1688, "moment_max_abs_diff": 0.00012739747762680054, "step": 211, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00013790279626846313, "loss_diff": 1.8596649169921875e-05, "max_param_diff": 0.003133709542453289, "microstep": 1696, "moment_max_abs_diff": 0.0001284480094909668, "step": 212, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017104297876358032, "loss_diff": 1.6689300537109375e-06, "max_param_diff": 0.003134281374514103, "microstep": 1704, "moment_max_abs_diff": 0.00013270974159240723, "step": 213, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017699599266052246, "loss_diff": 5.9604644775390625e-06, "max_param_diff": 0.0031347963958978653, "microstep": 1712, "moment_max_abs_diff": 0.00013713911175727844, "step": 214, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00014513731002807617, "loss_diff": 8.106231689453125e-06, "max_param_diff": 0.003135260194540024, "microstep": 1720, "moment_max_abs_diff": 0.00013794004917144775, "step": 215, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00013300776481628418, "loss_diff": 5.245208740234375e-06, "max_param_diff": 0.0031356774270534515, "microstep": 1728, "moment_max_abs_diff": 0.00013744831085205078, "step": 216, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00014659250155091286, "loss_diff": 2.0503997802734375e-05, "max_param_diff": 0.0031360527500510216, "microstep": 1736, "moment_max_abs_diff": 0.00012350454926490784, "step": 217, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001908615231513977, "loss_diff": 1.7881393432617188e-05, "max_param_diff": 0.0031363898888230324, "microstep": 1744, "moment_max_abs_diff": 0.00012708082795143127, "step": 218, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017329305410385132, "loss_diff": 1.4781951904296875e-05, "max_param_diff": 0.0031366944313049316, "microstep": 1752, "moment_max_abs_diff": 0.0001317039132118225, "step": 219, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00019241496920585632, "loss_diff": 1.3113021850585938e-05, "max_param_diff": 0.0031369682401418686, "microstep": 1760, "moment_max_abs_diff": 0.0001285932958126068, "step": 220, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000284343957901001, "loss_diff": 2.6941299438476562e-05, "max_param_diff": 0.0031372131779789925, "microstep": 1768, "moment_max_abs_diff": 0.00013167411088943481, "step": 221, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018325448036193848, "loss_diff": 5.4836273193359375e-06, "max_param_diff": 0.0031374339014291763, "microstep": 1776, "moment_max_abs_diff": 0.00012893974781036377, "step": 222, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015475228428840637, "loss_diff": 1.8358230590820312e-05, "max_param_diff": 0.0031376322731375694, "microstep": 1784, "moment_max_abs_diff": 0.00012801587581634521, "step": 223, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001690387725830078, "loss_diff": 2.5510787963867188e-05, "max_param_diff": 0.003137810155749321, "microstep": 1792, "moment_max_abs_diff": 0.0001305416226387024, "step": 224, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001840493641793728, "loss_diff": 1.7881393432617188e-05, "max_param_diff": 0.003137970343232155, "microstep": 1800, "moment_max_abs_diff": 0.00013151764869689941, "step": 225, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001467391848564148, "loss_diff": 1.6689300537109375e-06, "max_param_diff": 0.0031381137669086456, "microstep": 1808, "moment_max_abs_diff": 0.00013138353824615479, "step": 226, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018829107284545898, "loss_diff": 2.5272369384765625e-05, "max_param_diff": 0.0031382404267787933, "microstep": 1816, "moment_max_abs_diff": 0.0001283884048461914, "step": 227, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004098154604434967, "loss_diff": 3.910064697265625e-05, "max_param_diff": 0.0031383540481328964, "microstep": 1824, "moment_max_abs_diff": 0.00012244656682014465, "step": 228, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00018196599557995796, "loss_diff": 2.384185791015625e-06, "max_param_diff": 0.003138456493616104, "microstep": 1832, "moment_max_abs_diff": 0.000123538076877594, "step": 229, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00019621429964900017, "loss_diff": 1.6927719116210938e-05, "max_param_diff": 0.0031385477632284164, "microstep": 1840, "moment_max_abs_diff": 0.00012372061610221863, "step": 230, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00013079866766929626, "loss_diff": 2.765655517578125e-05, "max_param_diff": 0.0031386278569698334, "microstep": 1848, "moment_max_abs_diff": 0.00012422725558280945, "step": 231, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00013038143515586853, "loss_diff": 1.6927719116210938e-05, "max_param_diff": 0.0031387005001306534, "microstep": 1856, "moment_max_abs_diff": 0.00012484565377235413, "step": 232, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00014540553092956543, "loss_diff": 5.0067901611328125e-06, "max_param_diff": 0.0031387638300657272, "microstep": 1864, "moment_max_abs_diff": 0.0001255534589290619, "step": 233, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00011201947927474976, "loss_diff": 1.8358230590820312e-05, "max_param_diff": 0.003138819709420204, "microstep": 1872, "moment_max_abs_diff": 0.00012420117855072021, "step": 234, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000247269868850708, "loss_diff": 7.62939453125e-06, "max_param_diff": 0.003138868138194084, "microstep": 1880, "moment_max_abs_diff": 0.00011207908391952515, "step": 235, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003921389579772949, "loss_diff": 0.00013709068298339844, "max_param_diff": 0.003138914704322815, "microstep": 1888, "moment_max_abs_diff": 9.587034583091736e-05, "step": 236, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00021904613822698593, "loss_diff": 3.743171691894531e-05, "max_param_diff": 0.003138953819870949, "microstep": 1896, "moment_max_abs_diff": 8.0890953540802e-05, "step": 237, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020238757133483887, "loss_diff": 6.4373016357421875e-06, "max_param_diff": 0.003138989210128784, "microstep": 1904, "moment_max_abs_diff": 8.140876889228821e-05, "step": 238, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00038572773337364197, "loss_diff": 6.532669067382812e-05, "max_param_diff": 0.003139020875096321, "microstep": 1912, "moment_max_abs_diff": 8.877739310264587e-05, "step": 239, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002913982607424259, "loss_diff": 7.176399230957031e-05, "max_param_diff": 0.0031390488147735596, "microstep": 1920, "moment_max_abs_diff": 8.81049782037735e-05, "step": 240, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00038380175828933716, "loss_diff": 1.621246337890625e-05, "max_param_diff": 0.0031390730291604996, "microstep": 1928, "moment_max_abs_diff": 0.00010268948972225189, "step": 241, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002827644348144531, "loss_diff": 5.221366882324219e-05, "max_param_diff": 0.003139093518257141, "microstep": 1936, "moment_max_abs_diff": 0.00010290183126926422, "step": 242, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002826545387506485, "loss_diff": 1.049041748046875e-05, "max_param_diff": 0.003139110282063484, "microstep": 1944, "moment_max_abs_diff": 9.982846677303314e-05, "step": 243, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00036475062370300293, "loss_diff": 2.4557113647460938e-05, "max_param_diff": 0.003139125183224678, "microstep": 1952, "moment_max_abs_diff": 8.691288530826569e-05, "step": 244, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00024382397532463074, "loss_diff": 5.0067901611328125e-05, "max_param_diff": 0.0031391382217407227, "microstep": 1960, "moment_max_abs_diff": 8.995458483695984e-05, "step": 245, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003700805827975273, "loss_diff": 3.814697265625e-05, "max_param_diff": 0.003139149397611618, "microstep": 1968, "moment_max_abs_diff": 9.439513087272644e-05, "step": 246, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00022600684314966202, "loss_diff": 3.695487976074219e-05, "max_param_diff": 0.003139156848192215, "microstep": 1976, "moment_max_abs_diff": 0.00010069645941257477, "step": 247, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003272220492362976, "loss_diff": 2.6226043701171875e-06, "max_param_diff": 0.003139164298772812, "microstep": 1984, "moment_max_abs_diff": 0.00010265596210956573, "step": 248, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00019820407032966614, "loss_diff": 1.8835067749023438e-05, "max_param_diff": 0.0031391698867082596, "microstep": 1992, "moment_max_abs_diff": 0.0001019556075334549, "step": 249, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001763487234711647, "loss_diff": 3.933906555175781e-05, "max_param_diff": 0.003139173611998558, "microstep": 2000, "moment_max_abs_diff": 0.00010665133595466614, "step": 250, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004666075110435486, "loss_diff": 1.7404556274414062e-05, "max_param_diff": 0.0031391754746437073, "microstep": 2008, "moment_max_abs_diff": 0.00010629743337631226, "step": 251, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020470470190048218, "loss_diff": 2.384185791015625e-06, "max_param_diff": 0.003110591322183609, "microstep": 2016, "moment_max_abs_diff": 0.00010944157838821411, "step": 252, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003536529839038849, "loss_diff": 6.270408630371094e-05, "max_param_diff": 0.003084808588027954, "microstep": 2024, "moment_max_abs_diff": 0.00010697543621063232, "step": 253, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002861861139535904, "loss_diff": 3.1948089599609375e-05, "max_param_diff": 0.003061549738049507, "microstep": 2032, "moment_max_abs_diff": 0.00010515749454498291, "step": 254, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00047088414430618286, "loss_diff": 2.5033950805664062e-05, "max_param_diff": 0.003040570765733719, "microstep": 2040, "moment_max_abs_diff": 0.00010240823030471802, "step": 255, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004926901310682297, "loss_diff": 2.6941299438476562e-05, "max_param_diff": 0.0030216462910175323, "microstep": 2048, "moment_max_abs_diff": 9.130686521530151e-05, "step": 256, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00033880025148391724, "loss_diff": 5.0067901611328125e-06, "max_param_diff": 0.00300457701086998, "microstep": 2056, "moment_max_abs_diff": 0.00010016933083534241, "step": 257, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005625709891319275, "loss_diff": 2.8371810913085938e-05, "max_param_diff": 0.0029891785234212875, "microstep": 2064, "moment_max_abs_diff": 0.00010637938976287842, "step": 258, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00023108720779418945, "loss_diff": 1.6689300537109375e-05, "max_param_diff": 0.0029752906411886215, "microstep": 2072, "moment_max_abs_diff": 0.00010646507143974304, "step": 259, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00042578577995300293, "loss_diff": 8.344650268554688e-06, "max_param_diff": 0.002962762489914894, "microstep": 2080, "moment_max_abs_diff": 0.00010637566447257996, "step": 260, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002990700304508209, "loss_diff": 4.267692565917969e-05, "max_param_diff": 0.00295146182179451, "microstep": 2088, "moment_max_abs_diff": 0.00010325759649276733, "step": 261, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00022049620747566223, "loss_diff": 3.790855407714844e-05, "max_param_diff": 0.002941267564892769, "microstep": 2096, "moment_max_abs_diff": 0.00010339915752410889, "step": 262, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00019471347332000732, "loss_diff": 5.53131103515625e-05, "max_param_diff": 0.0029320716857910156, "microstep": 2104, "moment_max_abs_diff": 0.00010719150304794312, "step": 263, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0001536533236503601, "loss_diff": 3.504753112792969e-05, "max_param_diff": 0.00292377732694149, "microstep": 2112, "moment_max_abs_diff": 0.00011040642857551575, "step": 264, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00017579179257154465, "loss_diff": 3.838539123535156e-05, "max_param_diff": 0.0029162950813770294, "microstep": 2120, "moment_max_abs_diff": 0.00011077523231506348, "step": 265, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00033301860094070435, "loss_diff": 4.2438507080078125e-05, "max_param_diff": 0.002909542992711067, "microstep": 2128, "moment_max_abs_diff": 0.00010846555233001709, "step": 266, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00043817050755023956, "loss_diff": 2.1457672119140625e-06, "max_param_diff": 0.002903454005718231, "microstep": 2136, "moment_max_abs_diff": 0.00010434165596961975, "step": 267, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00027624890208244324, "loss_diff": 4.744529724121094e-05, "max_param_diff": 0.002897959202528, "microstep": 2144, "moment_max_abs_diff": 0.00010506808757781982, "step": 268, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00015842542052268982, "loss_diff": 4.1484832763671875e-05, "max_param_diff": 0.0028930045664310455, "microstep": 2152, "moment_max_abs_diff": 0.00010916218161582947, "step": 269, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00028637051582336426, "loss_diff": 4.76837158203125e-05, "max_param_diff": 0.0028885342180728912, "microstep": 2160, "moment_max_abs_diff": 0.00010924413800239563, "step": 270, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00016766414046287537, "loss_diff": 1.3828277587890625e-05, "max_param_diff": 0.0028845034539699554, "microstep": 2168, "moment_max_abs_diff": 0.00011192262172698975, "step": 271, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00023810938000679016, "loss_diff": 3.337860107421875e-05, "max_param_diff": 0.0028808657079935074, "microstep": 2176, "moment_max_abs_diff": 0.00011373311281204224, "step": 272, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002100728452205658, "loss_diff": 5.9604644775390625e-05, "max_param_diff": 0.0028775837272405624, "microstep": 2184, "moment_max_abs_diff": 0.0001147240400314331, "step": 273, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004718024283647537, "loss_diff": 9.298324584960938e-05, "max_param_diff": 0.0028746239840984344, "microstep": 2192, "moment_max_abs_diff": 0.00010717660188674927, "step": 274, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006144056096673012, "loss_diff": 6.079673767089844e-05, "max_param_diff": 0.0028719529509544373, "microstep": 2200, "moment_max_abs_diff": 9.277649223804474e-05, "step": 275, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003335997462272644, "loss_diff": 0.00011801719665527344, "max_param_diff": 0.0028695426881313324, "microstep": 2208, "moment_max_abs_diff": 0.00010975264012813568, "step": 276, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005352944135665894, "loss_diff": 2.6702880859375e-05, "max_param_diff": 0.00286736898124218, "microstep": 2216, "moment_max_abs_diff": 0.0001249350607395172, "step": 277, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004902109503746033, "loss_diff": 3.504753112792969e-05, "max_param_diff": 0.0028654057532548904, "microstep": 2224, "moment_max_abs_diff": 8.506886661052704e-05, "step": 278, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003822571597993374, "loss_diff": 3.2901763916015625e-05, "max_param_diff": 0.002863634377717972, "microstep": 2232, "moment_max_abs_diff": 8.301436901092529e-05, "step": 279, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00034268107265233994, "loss_diff": 8.726119995117188e-05, "max_param_diff": 0.0028620362281799316, "microstep": 2240, "moment_max_abs_diff": 9.335484355688095e-05, "step": 280, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00044033676385879517, "loss_diff": 5.1975250244140625e-05, "max_param_diff": 0.002860594540834427, "microstep": 2248, "moment_max_abs_diff": 0.00010142382234334946, "step": 281, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00046303868293762207, "loss_diff": 2.0265579223632812e-05, "max_param_diff": 0.0028592944145202637, "microstep": 2256, "moment_max_abs_diff": 0.00010017026215791702, "step": 282, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006119962781667709, "loss_diff": 7.963180541992188e-05, "max_param_diff": 0.002858120948076248, "microstep": 2264, "moment_max_abs_diff": 8.765794336795807e-05, "step": 283, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00034424662590026855, "loss_diff": 0.00016164779663085938, "max_param_diff": 0.0028570611029863358, "microstep": 2272, "moment_max_abs_diff": 8.813804015517235e-05, "step": 284, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005628243088722229, "loss_diff": 5.841255187988281e-05, "max_param_diff": 0.002856103703379631, "microstep": 2280, "moment_max_abs_diff": 8.848821744322777e-05, "step": 285, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004033297300338745, "loss_diff": 5.793571472167969e-05, "max_param_diff": 0.002855241298675537, "microstep": 2288, "moment_max_abs_diff": 9.359186515212059e-05, "step": 286, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004483237862586975, "loss_diff": 1.71661376953125e-05, "max_param_diff": 0.0028544608503580093, "microstep": 2296, "moment_max_abs_diff": 0.0001108907163143158, "step": 287, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00020456314086914062, "loss_diff": 5.936622619628906e-05, "max_param_diff": 0.0028537567704916, "microstep": 2304, "moment_max_abs_diff": 9.326846338808537e-05, "step": 288, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004764571785926819, "loss_diff": 8.106231689453125e-05, "max_param_diff": 0.002853119745850563, "microstep": 2312, "moment_max_abs_diff": 9.162351489067078e-05, "step": 289, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006515122950077057, "loss_diff": 2.7418136596679688e-05, "max_param_diff": 0.0028525460511446, "microstep": 2320, "moment_max_abs_diff": 0.00010444736108183861, "step": 290, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00036054477095603943, "loss_diff": 2.5510787963867188e-05, "max_param_diff": 0.002823389135301113, "microstep": 2328, "moment_max_abs_diff": 0.00012336019426584244, "step": 291, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00039504654705524445, "loss_diff": 5.078315734863281e-05, "max_param_diff": 0.002797096036374569, "microstep": 2336, "moment_max_abs_diff": 0.00011842884123325348, "step": 292, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004509463906288147, "loss_diff": 8.463859558105469e-05, "max_param_diff": 0.002773384563624859, "microstep": 2344, "moment_max_abs_diff": 9.153597056865692e-05, "step": 293, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0008730366826057434, "loss_diff": 1.9550323486328125e-05, "max_param_diff": 0.002752002328634262, "microstep": 2352, "moment_max_abs_diff": 8.293241262435913e-05, "step": 294, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000724300742149353, "loss_diff": 6.651878356933594e-05, "max_param_diff": 0.0027327202260494232, "microstep": 2360, "moment_max_abs_diff": 9.328452870249748e-05, "step": 295, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00035322830080986023, "loss_diff": 5.245208740234375e-05, "max_param_diff": 0.0027153315022587776, "microstep": 2368, "moment_max_abs_diff": 9.15229320526123e-05, "step": 296, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0007209479808807373, "loss_diff": 3.0517578125e-05, "max_param_diff": 0.002699650824069977, "microstep": 2376, "moment_max_abs_diff": 9.836256504058838e-05, "step": 297, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005486756563186646, "loss_diff": 6.771087646484375e-05, "max_param_diff": 0.0026855096220970154, "microstep": 2384, "moment_max_abs_diff": 0.00010216608643531799, "step": 298, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004359334707260132, "loss_diff": 8.58306884765625e-06, "max_param_diff": 0.0026727570220828056, "microstep": 2392, "moment_max_abs_diff": 0.0001025274395942688, "step": 299, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00042368844151496887, "loss_diff": 1.239776611328125e-05, "max_param_diff": 0.002661256119608879, "microstep": 2400, "moment_max_abs_diff": 0.00010089948773384094, "step": 300, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00032846443355083466, "loss_diff": 5.2928924560546875e-05, "max_param_diff": 0.002650884911417961, "microstep": 2408, "moment_max_abs_diff": 9.728595614433289e-05, "step": 301, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004132399335503578, "loss_diff": 4.57763671875e-05, "max_param_diff": 0.0026415325701236725, "microstep": 2416, "moment_max_abs_diff": 9.856745600700378e-05, "step": 302, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002619698643684387, "loss_diff": 9.059906005859375e-06, "max_param_diff": 0.0026330985128879547, "microstep": 2424, "moment_max_abs_diff": 9.767338633537292e-05, "step": 303, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005507171154022217, "loss_diff": 1.2636184692382812e-05, "max_param_diff": 0.0026254933327436447, "microstep": 2432, "moment_max_abs_diff": 9.398534893989563e-05, "step": 304, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0003907773643732071, "loss_diff": 3.2901763916015625e-05, "max_param_diff": 0.0026186322793364525, "microstep": 2440, "moment_max_abs_diff": 9.256601333618164e-05, "step": 305, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005610287189483643, "loss_diff": 9.72747802734375e-05, "max_param_diff": 0.0026124464347958565, "microstep": 2448, "moment_max_abs_diff": 9.29906964302063e-05, "step": 306, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004305727779865265, "loss_diff": 3.933906555175781e-05, "max_param_diff": 0.0026068678125739098, "microstep": 2456, "moment_max_abs_diff": 9.262189269065857e-05, "step": 307, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002644956111907959, "loss_diff": 5.221366882324219e-05, "max_param_diff": 0.002601836808025837, "microstep": 2464, "moment_max_abs_diff": 0.00010004639625549316, "step": 308, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006033331155776978, "loss_diff": 3.6716461181640625e-05, "max_param_diff": 0.0025972984731197357, "microstep": 2472, "moment_max_abs_diff": 0.00010848790407180786, "step": 309, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0002861209213733673, "loss_diff": 0.00011014938354492188, "max_param_diff": 0.00259320717304945, "microstep": 2480, "moment_max_abs_diff": 0.00011586956679821014, "step": 310, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005873292684555054, "loss_diff": 3.933906555175781e-05, "max_param_diff": 0.0025895163416862488, "microstep": 2488, "moment_max_abs_diff": 0.0001222752034664154, "step": 311, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00026960670948028564, "loss_diff": 5.1975250244140625e-05, "max_param_diff": 0.002586187794804573, "microstep": 2496, "moment_max_abs_diff": 0.00012045353651046753, "step": 312, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00073291826993227, "loss_diff": 5.245208740234375e-05, "max_param_diff": 0.0025831861421465874, "microstep": 2504, "moment_max_abs_diff": 0.00011442974209785461, "step": 313, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000729680061340332, "loss_diff": 6.341934204101562e-05, "max_param_diff": 0.00258047878742218, "microstep": 2512, "moment_max_abs_diff": 0.00010062102228403091, "step": 314, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004685390740633011, "loss_diff": 9.131431579589844e-05, "max_param_diff": 0.0025780368596315384, "microstep": 2520, "moment_max_abs_diff": 0.0001260600984096527, "step": 315, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006366372108459473, "loss_diff": 8.511543273925781e-05, "max_param_diff": 0.002575834281742573, "microstep": 2528, "moment_max_abs_diff": 0.00017051398754119873, "step": 316, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000573744997382164, "loss_diff": 2.384185791015625e-07, "max_param_diff": 0.002573847770690918, "microstep": 2536, "moment_max_abs_diff": 0.00011748261749744415, "step": 317, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0008504539728164673, "loss_diff": 0.00011014938354492188, "max_param_diff": 0.002572055906057358, "microstep": 2544, "moment_max_abs_diff": 0.00010830909013748169, "step": 318, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004722326993942261, "loss_diff": 6.0558319091796875e-05, "max_param_diff": 0.0025704391300678253, "microstep": 2552, "moment_max_abs_diff": 0.00010946020483970642, "step": 319, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000635191798210144, "loss_diff": 0.00014829635620117188, "max_param_diff": 0.0025689806789159775, "microstep": 2560, "moment_max_abs_diff": 0.00010734423995018005, "step": 320, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0007930248975753784, "loss_diff": 0.00010538101196289062, "max_param_diff": 0.002567664720118046, "microstep": 2568, "moment_max_abs_diff": 0.00012494437396526337, "step": 321, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.001260131597518921, "loss_diff": 3.814697265625e-06, "max_param_diff": 0.002566477283835411, "microstep": 2576, "moment_max_abs_diff": 0.00013713352382183075, "step": 322, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006433427333831787, "loss_diff": 2.0742416381835938e-05, "max_param_diff": 0.0025654062628746033, "microstep": 2584, "moment_max_abs_diff": 0.00013622362166643143, "step": 323, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006649643182754517, "loss_diff": 5.698204040527344e-05, "max_param_diff": 0.002564438618719578, "microstep": 2592, "moment_max_abs_diff": 0.00012596696615219116, "step": 324, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005195401608943939, "loss_diff": 8.678436279296875e-05, "max_param_diff": 0.0025635669007897377, "microstep": 2600, "moment_max_abs_diff": 0.000141829252243042, "step": 325, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004792846739292145, "loss_diff": 9.369850158691406e-05, "max_param_diff": 0.0025627799332141876, "microstep": 2608, "moment_max_abs_diff": 0.00012804940342903137, "step": 326, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005642995238304138, "loss_diff": 5.698204040527344e-05, "max_param_diff": 0.002562069334089756, "microstep": 2616, "moment_max_abs_diff": 0.0001118960790336132, "step": 327, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0008697826415300369, "loss_diff": 5.0067901611328125e-05, "max_param_diff": 0.002561427652835846, "microstep": 2624, "moment_max_abs_diff": 0.00011546770110726357, "step": 328, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0007066130638122559, "loss_diff": 0.00010657310485839844, "max_param_diff": 0.002560848370194435, "microstep": 2632, "moment_max_abs_diff": 0.00014419108629226685, "step": 329, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005969163030385971, "loss_diff": 2.193450927734375e-05, "max_param_diff": 0.002560325898230076, "microstep": 2640, "moment_max_abs_diff": 0.00016854330897331238, "step": 330, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.00043831299990415573, "loss_diff": 6.628036499023438e-05, "max_param_diff": 0.0025577759370207787, "microstep": 2648, "moment_max_abs_diff": 0.0001795068383216858, "step": 331, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0009704232215881348, "loss_diff": 2.2411346435546875e-05, "max_param_diff": 0.002555476501584053, "microstep": 2656, "moment_max_abs_diff": 0.0001243147999048233, "step": 332, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0010524839162826538, "loss_diff": 1.621246337890625e-05, "max_param_diff": 0.0025534024462103844, "microstep": 2664, "moment_max_abs_diff": 0.00014471868053078651, "step": 333, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0007569137960672379, "loss_diff": 5.6743621826171875e-05, "max_param_diff": 0.0025515323504805565, "microstep": 2672, "moment_max_abs_diff": 0.00018296297639608383, "step": 334, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005934685468673706, "loss_diff": 3.910064697265625e-05, "max_param_diff": 0.0025498447939753532, "microstep": 2680, "moment_max_abs_diff": 0.0001619141548871994, "step": 335, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0009158924221992493, "loss_diff": 0.00010323524475097656, "max_param_diff": 0.0025483230128884315, "microstep": 2688, "moment_max_abs_diff": 0.00011183694005012512, "step": 336, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0009603649377822876, "loss_diff": 4.8160552978515625e-05, "max_param_diff": 0.0025469493120908737, "microstep": 2696, "moment_max_abs_diff": 0.00011360272765159607, "step": 337, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005929470062255859, "loss_diff": 4.935264587402344e-05, "max_param_diff": 0.0025457097217440605, "microstep": 2704, "moment_max_abs_diff": 0.00012785755097866058, "step": 338, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000841975212097168, "loss_diff": 6.008148193359375e-05, "max_param_diff": 0.0025445930659770966, "microstep": 2712, "moment_max_abs_diff": 0.00014428235590457916, "step": 339, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005059130489826202, "loss_diff": 0.00015592575073242188, "max_param_diff": 0.002543584443628788, "microstep": 2720, "moment_max_abs_diff": 0.00016331486403942108, "step": 340, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004645492881536484, "loss_diff": 3.9577484130859375e-05, "max_param_diff": 0.0025426745414733887, "microstep": 2728, "moment_max_abs_diff": 0.00014493148773908615, "step": 341, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0004872828722000122, "loss_diff": 3.409385681152344e-05, "max_param_diff": 0.002541853114962578, "microstep": 2736, "moment_max_abs_diff": 0.00010468624532222748, "step": 342, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0007408503443002701, "loss_diff": 5.078315734863281e-05, "max_param_diff": 0.002541111782193184, "microstep": 2744, "moment_max_abs_diff": 9.944848716259003e-05, "step": 343, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000858839601278305, "loss_diff": 9.417533874511719e-05, "max_param_diff": 0.0025404421612620354, "microstep": 2752, "moment_max_abs_diff": 0.00013811327517032623, "step": 344, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0008276328444480896, "loss_diff": 0.0003833770751953125, "max_param_diff": 0.0025398386642336845, "microstep": 2760, "moment_max_abs_diff": 0.00019907206296920776, "step": 345, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006753802299499512, "loss_diff": 4.363059997558594e-05, "max_param_diff": 0.0025392938405275345, "microstep": 2768, "moment_max_abs_diff": 0.00022853165864944458, "step": 346, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0007544457912445068, "loss_diff": 0.00011658668518066406, "max_param_diff": 0.002538801170885563, "microstep": 2776, "moment_max_abs_diff": 0.00020135845988988876, "step": 347, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006073154509067535, "loss_diff": 9.775161743164062e-06, "max_param_diff": 0.0025383559986948967, "microstep": 2784, "moment_max_abs_diff": 0.00014546886086463928, "step": 348, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005561783909797668, "loss_diff": 6.175041198730469e-05, "max_param_diff": 0.0025379545986652374, "microstep": 2792, "moment_max_abs_diff": 0.00012367591261863708, "step": 349, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0007108114659786224, "loss_diff": 0.00011801719665527344, "max_param_diff": 0.002537592314183712, "microstep": 2800, "moment_max_abs_diff": 0.00014963559806346893, "step": 350, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0005478388629853725, "loss_diff": 6.151199340820312e-05, "max_param_diff": 0.0025372635573148727, "microstep": 2808, "moment_max_abs_diff": 0.00013362709432840347, "step": 351, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.001973360776901245, "loss_diff": 8.034706115722656e-05, "max_param_diff": 0.002536967396736145, "microstep": 2816, "moment_max_abs_diff": 0.00013076141476631165, "step": 352, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0015470758080482483, "loss_diff": 0.00015401840209960938, "max_param_diff": 0.0025366991758346558, "microstep": 2824, "moment_max_abs_diff": 0.0001635603839531541, "step": 353, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0016598217189311981, "loss_diff": 9.083747863769531e-05, "max_param_diff": 0.0025364570319652557, "microstep": 2832, "moment_max_abs_diff": 0.00020477641373872757, "step": 354, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0010730288922786713, "loss_diff": 5.53131103515625e-05, "max_param_diff": 0.002536238171160221, "microstep": 2840, "moment_max_abs_diff": 0.00022830264060758054, "step": 355, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0008672922849655151, "loss_diff": 7.987022399902344e-05, "max_param_diff": 0.002536039799451828, "microstep": 2848, "moment_max_abs_diff": 0.0002269931137561798, "step": 356, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0008095130324363708, "loss_diff": 6.198883056640625e-06, "max_param_diff": 0.002535860054194927, "microstep": 2856, "moment_max_abs_diff": 0.00019760243594646454, "step": 357, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0012424327433109283, "loss_diff": 5.0067901611328125e-05, "max_param_diff": 0.002535698004066944, "microstep": 2864, "moment_max_abs_diff": 0.00019354280084371567, "step": 358, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0012231171131134033, "loss_diff": 0.00014495849609375, "max_param_diff": 0.002535550855100155, "microstep": 2872, "moment_max_abs_diff": 0.0002065412700176239, "step": 359, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0013401694595813751, "loss_diff": 0.00011467933654785156, "max_param_diff": 0.002535417675971985, "microstep": 2880, "moment_max_abs_diff": 0.00024070573272183537, "step": 360, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0016986280679702759, "loss_diff": 0.00012373924255371094, "max_param_diff": 0.002535296604037285, "microstep": 2888, "moment_max_abs_diff": 0.0002460433170199394, "step": 361, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0013226782903075218, "loss_diff": 3.9577484130859375e-05, "max_param_diff": 0.0025351857766509056, "microstep": 2896, "moment_max_abs_diff": 0.00026511261239647865, "step": 362, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0010062344372272491, "loss_diff": 8.606910705566406e-05, "max_param_diff": 0.002535085193812847, "microstep": 2904, "moment_max_abs_diff": 0.00017505139112472534, "step": 363, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.001311885192990303, "loss_diff": 1.3828277587890625e-05, "max_param_diff": 0.0025349948555231094, "microstep": 2912, "moment_max_abs_diff": 0.00025325827300548553, "step": 364, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006336607038974762, "loss_diff": 4.649162292480469e-05, "max_param_diff": 0.0025349119678139687, "microstep": 2920, "moment_max_abs_diff": 0.00022747227922081947, "step": 365, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.000882267951965332, "loss_diff": 7.748603820800781e-05, "max_param_diff": 0.002534836530685425, "microstep": 2928, "moment_max_abs_diff": 0.00020477361977100372, "step": 366, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0014449357986450195, "loss_diff": 7.05718994140625e-05, "max_param_diff": 0.002534768544137478, "microstep": 2936, "moment_max_abs_diff": 0.0001810230314731598, "step": 367, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0011337623000144958, "loss_diff": 4.982948303222656e-05, "max_param_diff": 0.0025347061455249786, "microstep": 2944, "moment_max_abs_diff": 0.00016320962458848953, "step": 368, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0008397847414016724, "loss_diff": 0.00012159347534179688, "max_param_diff": 0.0025346484035253525, "microstep": 2952, "moment_max_abs_diff": 0.00021460093557834625, "step": 369, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0011339876800775528, "loss_diff": 7.867813110351562e-05, "max_param_diff": 0.002521177753806114, "microstep": 2960, "moment_max_abs_diff": 0.00028454139828681946, "step": 370, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0006006471812725067, "loss_diff": 0.00016236305236816406, "max_param_diff": 0.0025090333074331284, "microstep": 2968, "moment_max_abs_diff": 0.000261601060628891, "step": 371, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0009581446647644043, "loss_diff": 0.0001780986785888672, "max_param_diff": 0.0024980856105685234, "microstep": 2976, "moment_max_abs_diff": 0.00025248900055885315, "step": 372, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0013543516397476196, "loss_diff": 4.124641418457031e-05, "max_param_diff": 0.002488215919584036, "microstep": 2984, "moment_max_abs_diff": 0.00015791412442922592, "step": 373, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0018565990030765533, "loss_diff": 5.793571472167969e-05, "max_param_diff": 0.0024793175980448723, "microstep": 2992, "moment_max_abs_diff": 0.00015362165868282318, "step": 374, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0019222348928451538, "loss_diff": 3.218650817871094e-05, "max_param_diff": 0.002471296116709709, "microstep": 3000, "moment_max_abs_diff": 0.00028495490550994873, "step": 375, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0012745223939418793, "loss_diff": 8.535385131835938e-05, "max_param_diff": 0.0024640639312565327, "microstep": 3008, "moment_max_abs_diff": 0.00037949439138174057, "step": 376, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0008459985256195068, "loss_diff": 2.8848648071289062e-05, "max_param_diff": 0.0024575437419116497, "microstep": 3016, "moment_max_abs_diff": 0.00027621444314718246, "step": 377, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0015563666820526123, "loss_diff": 0.0002295970916748047, "max_param_diff": 0.0024516661651432514, "microstep": 3024, "moment_max_abs_diff": 0.00026219896972179413, "step": 378, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0026421844959259033, "loss_diff": 0.00017499923706054688, "max_param_diff": 0.002446366474032402, "microstep": 3032, "moment_max_abs_diff": 0.00021294038742780685, "step": 379, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0015422701835632324, "loss_diff": 4.506111145019531e-05, "max_param_diff": 0.002441589254885912, "microstep": 3040, "moment_max_abs_diff": 0.00028459541499614716, "step": 380, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0013911426067352295, "loss_diff": 3.0994415283203125e-05, "max_param_diff": 0.0024372823536396027, "microstep": 3048, "moment_max_abs_diff": 0.0003605126403272152, "step": 381, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0018316209316253662, "loss_diff": 7.843971252441406e-05, "max_param_diff": 0.002437267452478409, "microstep": 3056, "moment_max_abs_diff": 0.00021938607096672058, "step": 382, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0018900483846664429, "loss_diff": 2.9325485229492188e-05, "max_param_diff": 0.002437487244606018, "microstep": 3064, "moment_max_abs_diff": 0.00025077536702156067, "step": 383, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0016551762819290161, "loss_diff": 0.00011587142944335938, "max_param_diff": 0.0024376846849918365, "microstep": 3072, "moment_max_abs_diff": 0.000253134872764349, "step": 384, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.002266019582748413, "loss_diff": 8.940696716308594e-05, "max_param_diff": 0.0024378616362810135, "microstep": 3080, "moment_max_abs_diff": 0.0003429800271987915, "step": 385, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.001352548599243164, "loss_diff": 5.125999450683594e-05, "max_param_diff": 0.002438019961118698, "microstep": 3088, "moment_max_abs_diff": 0.0003159581683576107, "step": 386, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0015263594686985016, "loss_diff": 2.9802322387695312e-05, "max_param_diff": 0.0024381615221500397, "microstep": 3096, "moment_max_abs_diff": 0.00022194022312760353, "step": 387, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0014804080128669739, "loss_diff": 1.9311904907226562e-05, "max_param_diff": 0.0024382881820201874, "microstep": 3104, "moment_max_abs_diff": 0.0001790625974535942, "step": 388, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.002009153366088867, "loss_diff": 0.0002560615539550781, "max_param_diff": 0.0024384018033742905, "microstep": 3112, "moment_max_abs_diff": 0.00021583214402198792, "step": 389, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0023641586303710938, "loss_diff": 2.9325485229492188e-05, "max_param_diff": 0.002438504248857498, "microstep": 3120, "moment_max_abs_diff": 0.00027955789119005203, "step": 390, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0015805475413799286, "loss_diff": 0.0003190040588378906, "max_param_diff": 0.0024385955184698105, "microstep": 3128, "moment_max_abs_diff": 0.00031878240406513214, "step": 391, "worst_param": "reported_in_final_tensor_inventory" }, { "gradient_max_abs_diff": 0.0038646459579467773, "loss_diff": 0.0, "max_param_diff": 0.0024386756122112274, "microstep": 3130, "moment_max_abs_diff": 0.0004012342542409897, "partial_window": true, "step": 392, "worst_param": "reported_in_final_tensor_inventory" } ] }, "runtime_provenance": { "export_rebuild": { "actual_runtime_threads": 2, "build_identity": { "expected": "f6f6168bad4501ce8b8cc64b3d063cc694738117fb31599eeec9b814182c13c5", "loaded": "f6f6168bad4501ce8b8cc64b3d063cc694738117fb31599eeec9b814182c13c5", "passed": true }, "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525944, "inode": 9176195, "mtime_ns": 1789977703702379527, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" } ], "loaded_sha256": [ "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ], "observed_in_process_maps": true, "passed": true, "process_map_entries": [ "7f3b0e947000-7f3b0e95e000 r--p 00000000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f3b0e95e000-7f3b0ea6e000 r-xp 00017000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f3b0ea6e000-7f3b0eaa0000 r--p 00127000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f3b0eaa0000-7f3b0eaa4000 r--p 00159000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f3b0eaa4000-7f3b0eaa5000 rw-p 0015d000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ] }, "fresh_process": { "command": [ "/opt/hostedtoolcache/Python/3.11.16/x64/bin/python", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/scripts/run_training_certification_v8.py", "--probe-loaded-dependencies", "--probe-library", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "--probe-json-out", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/loaded_dependency_probe_export_rebuild.json", "--probe-runtime-contract", "a77ec93e2250511a28f721e6d7c7fa1e98fcd08ffd755d416f25a66de29728d2", "--probe-build-identity", "f6f6168bad4501ce8b8cc64b3d063cc694738117fb31599eeec9b814182c13c5", "--probe-engine-sha256", "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525944, "inode": 9176195, "mtime_ns": 1789977703702379527, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" } ], "loaded_sha256": [ "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ], "observed_in_process_maps": true, "passed": true, "process_map_entries": [ "7f68dbb7d000-7f68dbb94000 r--p 00000000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f68dbb94000-7f68dbca4000 r-xp 00017000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f68dbca4000-7f68dbcd6000 r--p 00127000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f68dbcd6000-7f68dbcda000 r--p 00159000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f68dbcda000-7f68dbcdb000 rw-p 0015d000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ] }, "passed": true, "returncode": 0, "stdout": "", "training_build_identity": { "expected": "f6f6168bad4501ce8b8cc64b3d063cc694738117fb31599eeec9b814182c13c5", "loaded": "f6f6168bad4501ce8b8cc64b3d063cc694738117fb31599eeec9b814182c13c5", "passed": true }, "training_contract": { "expected": "a77ec93e2250511a28f721e6d7c7fa1e98fcd08ffd755d416f25a66de29728d2", "loaded": "a77ec93e2250511a28f721e6d7c7fa1e98fcd08ffd755d416f25a66de29728d2", "passed": true }, "training_library": { "bytes": 332448, "inode": 9437798, "mtime_ns": 1789978806339383064, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "process_map_entries": [ "7f68e93b6000-7f68e93c0000 r--p 00000000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "7f68e93c0000-7f68e93f0000 r-xp 0000a000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "7f68e93f0000-7f68e93f7000 r--p 0003a000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "7f68e93f7000-7f68e93fe000 r--p 00041000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "7f68e93fe000-7f68e93ff000 rw-p 00048000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so" ], "sha256": "e8f6a224469dc71cd6deaff8b3d56a72cdaef205ee4a4e324c07a03559023931" } }, "passed": true, "role": "export_rebuild", "runtime_contract": { "expected": "a77ec93e2250511a28f721e6d7c7fa1e98fcd08ffd755d416f25a66de29728d2", "loaded": "a77ec93e2250511a28f721e6d7c7fa1e98fcd08ffd755d416f25a66de29728d2", "passed": true }, "training_library": { "bytes": 332448, "inode": 9437798, "mtime_ns": 1789978806339383064, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "process_map_entries": [ "7f3b0fa53000-7f3b0fa5d000 r--p 00000000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "7f3b0fa5d000-7f3b0fa8d000 r-xp 0000a000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "7f3b0fa8d000-7f3b0fa94000 r--p 0003a000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "7f3b0fa94000-7f3b0fa9b000 r--p 00041000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so", "7f3b0fa9b000-7f3b0fa9c000 rw-p 00048000 08:01 9437798 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/exported_inference/libtrain.so" ], "sha256": "e8f6a224469dc71cd6deaff8b3d56a72cdaef205ee4a4e324c07a03559023931" } }, "generation_probe": { "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525944, "inode": 9437826, "mtime_ns": 1789978867194324560, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" } ], "loaded_sha256": [ "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so" ], "observed_in_process_maps": true, "process_map_entries": [ "7f2baac7d000-7f2baac94000 r--p 00000000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "7f2baac94000-7f2baada4000 r-xp 00017000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "7f2baada4000-7f2baadd6000 r--p 00127000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "7f2baadd6000-7f2baadda000 r--p 00159000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "7f2baadda000-7f2baaddb000 rw-p 0015d000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so" ] }, "passed": true, "training_library": { "bytes": 222048, "inode": 9437828, "mtime_ns": 1789978872228946783, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "process_map_entries": [ "7f2bb85cd000-7f2bb85d3000 r--p 00000000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "7f2bb85d3000-7f2bb85f7000 r-xp 00006000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "7f2bb85f7000-7f2bb85fd000 r--p 0002a000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "7f2bb85fd000-7f2bb85fe000 r--p 0002f000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "7f2bb85fe000-7f2bb85ff000 rw-p 00030000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so" ], "sha256": "123b414016d149be78104139230be74a7877fddd0132f4fc76fd13dc28ce26ac" } }, "inference_probe": { "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525944, "inode": 9437826, "mtime_ns": 1789978867194324560, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" } ], "loaded_sha256": [ "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so" ], "observed_in_process_maps": true, "process_map_entries": [ "7fda16e7d000-7fda16e94000 r--p 00000000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "7fda16e94000-7fda16fa4000 r-xp 00017000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "7fda16fa4000-7fda16fd6000 r--p 00127000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "7fda16fd6000-7fda16fda000 r--p 00159000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so", "7fda16fda000-7fda16fdb000 rw-p 0015d000 08:01 9437826 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libckernel_engine.so" ] }, "passed": true, "training_library": { "bytes": 222048, "inode": 9437828, "mtime_ns": 1789978872228946783, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "process_map_entries": [ "7fda2475d000-7fda24763000 r--p 00000000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "7fda24763000-7fda24787000 r-xp 00006000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "7fda24787000-7fda2478d000 r--p 0002a000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "7fda2478d000-7fda2478e000 r--p 0002f000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so", "7fda2478e000-7fda2478f000 rw-p 00030000 08:01 9437828 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/v8_inference_runtime/libmodel.so" ], "sha256": "123b414016d149be78104139230be74a7877fddd0132f4fc76fd13dc28ce26ac" } }, "main_training": { "actual_runtime_threads": 2, "build_identity": { "expected": "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "loaded": "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "passed": true }, "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525944, "inode": 9176195, "mtime_ns": 1789977703702379527, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" } ], "loaded_sha256": [ "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ], "observed_in_process_maps": true, "passed": true, "process_map_entries": [ "7f3b0e947000-7f3b0e95e000 r--p 00000000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f3b0e95e000-7f3b0ea6e000 r-xp 00017000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f3b0ea6e000-7f3b0eaa0000 r--p 00127000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f3b0eaa0000-7f3b0eaa4000 r--p 00159000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f3b0eaa4000-7f3b0eaa5000 rw-p 0015d000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ] }, "fresh_process": { "command": [ "/opt/hostedtoolcache/Python/3.11.16/x64/bin/python", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/scripts/run_training_certification_v8.py", "--probe-loaded-dependencies", "--probe-library", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "--probe-json-out", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/loaded_dependency_probe_main_training.json", "--probe-runtime-contract", "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "--probe-build-identity", "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "--probe-engine-sha256", "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525944, "inode": 9176195, "mtime_ns": 1789977703702379527, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" } ], "loaded_sha256": [ "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ], "observed_in_process_maps": true, "passed": true, "process_map_entries": [ "7f1cf3b7d000-7f1cf3b94000 r--p 00000000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f1cf3b94000-7f1cf3ca4000 r-xp 00017000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f1cf3ca4000-7f1cf3cd6000 r--p 00127000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f1cf3cd6000-7f1cf3cda000 r--p 00159000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f1cf3cda000-7f1cf3cdb000 rw-p 0015d000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ] }, "passed": true, "returncode": 0, "stdout": "", "training_build_identity": { "expected": "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "loaded": "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "passed": true }, "training_contract": { "expected": "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "loaded": "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "passed": true }, "training_library": { "bytes": 332448, "inode": 9437765, "mtime_ns": 1789978764473586774, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "process_map_entries": [ "7f1cffab6000-7f1cffac0000 r--p 00000000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f1cffac0000-7f1cffaf0000 r-xp 0000a000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f1cffaf0000-7f1cffaf7000 r--p 0003a000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f1cffaf7000-7f1cffafe000 r--p 00041000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f1cffafe000-7f1cffaff000 rw-p 00048000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so" ], "sha256": "17561bede64a677b9a9e6b50f199fa7a1fdfde97547bece33fe45210d89019ec" } }, "passed": true, "role": "main_training", "runtime_contract": { "expected": "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "loaded": "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "passed": true }, "training_library": { "bytes": 332448, "inode": 9437765, "mtime_ns": 1789978764473586774, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "process_map_entries": [ "7f3b33eb6000-7f3b33ec0000 r--p 00000000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f3b33ec0000-7f3b33ef0000 r-xp 0000a000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f3b33ef0000-7f3b33ef7000 r--p 0003a000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f3b33ef7000-7f3b33efe000 r--p 00041000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f3b33efe000-7f3b33eff000 rw-p 00048000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so" ], "sha256": "17561bede64a677b9a9e6b50f199fa7a1fdfde97547bece33fe45210d89019ec" } }, "passed": true, "resume_worker": { "actual_runtime_threads": 2, "build_identity": { "expected": "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "loaded": "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "passed": true }, "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525944, "inode": 9176195, "mtime_ns": 1789977703702379527, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" } ], "loaded_sha256": [ "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ], "observed_in_process_maps": true, "passed": true, "process_map_entries": [ "7f57f9a7d000-7f57f9a94000 r--p 00000000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f57f9a94000-7f57f9ba4000 r-xp 00017000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f57f9ba4000-7f57f9bd6000 r--p 00127000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f57f9bd6000-7f57f9bda000 r--p 00159000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7f57f9bda000-7f57f9bdb000 rw-p 0015d000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ] }, "fresh_process": { "command": [ "/opt/hostedtoolcache/Python/3.11.16/x64/bin/python", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/scripts/run_training_certification_v8.py", "--probe-loaded-dependencies", "--probe-library", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "--probe-json-out", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/loaded_dependency_probe_resume_worker.json", "--probe-runtime-contract", "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "--probe-build-identity", "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "--probe-engine-sha256", "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "engine_dependency": { "basename": "libckernel_engine.so", "expected_sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a", "identities": [ { "bytes": 1525944, "inode": 9176195, "mtime_ns": 1789977703702379527, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "sha256": "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" } ], "loaded_sha256": [ "0eba76dd8f4cb95063d444202e9d6560503d79bc071b7da2fc72d3b12a77093a" ], "mapped_paths": [ "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ], "observed_in_process_maps": true, "passed": true, "process_map_entries": [ "7fb8cc67d000-7fb8cc694000 r--p 00000000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7fb8cc694000-7fb8cc7a4000 r-xp 00017000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7fb8cc7a4000-7fb8cc7d6000 r--p 00127000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7fb8cc7d6000-7fb8cc7da000 r--p 00159000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so", "7fb8cc7da000-7fb8cc7db000 rw-p 0015d000 08:01 9176195 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/build/libckernel_engine.so" ] }, "passed": true, "returncode": 0, "stdout": "", "training_build_identity": { "expected": "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "loaded": "3ce177edc1ca6773d2612e910ca05d46a75bfd0a03e0baf2522cbf91e42c51bb", "passed": true }, "training_contract": { "expected": "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "loaded": "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "passed": true }, "training_library": { "bytes": 332448, "inode": 9437765, "mtime_ns": 1789978764473586774, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "process_map_entries": [ "7fb8d84b6000-7fb8d84c0000 r--p 00000000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7fb8d84c0000-7fb8d84f0000 r-xp 0000a000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7fb8d84f0000-7fb8d84f7000 r--p 0003a000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7fb8d84f7000-7fb8d84fe000 r--p 00041000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7fb8d84fe000-7fb8d84ff000 rw-p 00048000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so" ], "sha256": "17561bede64a677b9a9e6b50f199fa7a1fdfde97547bece33fe45210d89019ec" } }, "passed": true, "role": "resume_worker", "runtime_contract": { "expected": "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "loaded": "fbb51760cbe4937699471cdf9583540ed2f163f243e01d2373e3cf3e4681f197", "passed": true }, "training_library": { "bytes": 332448, "inode": 9437765, "mtime_ns": 1789978764473586774, "observed_in_process_maps": true, "path": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "process_map_entries": [ "7f5807370000-7f580737a000 r--p 00000000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f580737a000-7f58073aa000 r-xp 0000a000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f58073aa000-7f58073b1000 r--p 0003a000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f58073b1000-7f58073b8000 r--p 00041000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so", "7f58073b8000-7f58073b9000 rw-p 00048000 08:01 9437765 /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/libtrain.so" ], "sha256": "17561bede64a677b9a9e6b50f199fa7a1fdfde97547bece33fe45210d89019ec" } } }, "training_ir_visualizer": { "artifacts": { "backprop_stitch_runtime_latest.json": "e1d8706d8bde9c48d911d5e108417fe2752a5cc0c451fdbcfbb885873a670b20", "dataset_profile.json": "26172897f0166966e1c042a9d4952672bfbdd0d69a622eda87b8c9aac7df0d05", "dataset_qc.json": "2f8d1fc4f3d789784adc72a230bf3c45b05b2dacbff6c1a88c3a41f21752126b", "tokenizer.json": "4f393c814420fb86101e7fe96bb9629857100b23d7aa5b90c6a9af1e8a3689a3", "tokenizer_roundtrip.json": "0193c9060b3c920dee831a1c5af6a50e250231f9f46025c1e7bf1ee0ccd42540", "training_checkpoint_policy_latest.json": "258031ec54b2eac30d510e6491b2607cccafab0dce3bcbf192307c2be9dd673c", "training_loss_curve_latest.json": "6c7a7ff6d2c25296599834bdf9b7a41e00b9a7813544271aeb3a4cec40ac3917", "training_parity_latest.json": "8fb89c2e429e4824819b1ec9dc56924738aed249c1062d823f5bb2db7a91ff42", "training_pipeline_latest.json": "ee11022ac4bb62dc7557b8627816979f1b6acebcaaa5bab78f1ff253851451a2", "training_step_profile_latest.json": "f9a08deb8d254b8fffa8d465ee9fcb2b09aa20da90ebf48184e41ebff0d5860c" }, "command": [ "/opt/hostedtoolcache/Python/3.11.16/x64/bin/python", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/tools/open_ir_visualizer_v8.py", "--generate", "--run", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32", "--html-only", "--strict-run-artifacts", "--output", "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/ir_report.html" ], "passed": true, "report": "/home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/ir_report.html", "report_sha256": "ecd0540ab7dfab229585e6f6bda95818425fc90b4fead04bf96ab167535a4027", "stdout": "Generating report for: qwen3_4layer_fp32\n - Missing optional files: ['memory_diagnostic', 'build_diagnostic', 'memory_verification', 'training_grad_norms', 'training_parity_regimen', 'training_parity_xray', 'training_canary_summary', 'run_ledger', 'corpus_sampling_log', 'post_train_eval', 'stage_eval_matrix', 'training_epoch_sweep', 'train_e2e', 'run_index', 'sanity_overfit', 'parity_report', 'profile_latest', 'contract_report', 'parity_1token', 'qk_norm_backward_parity', 'inference_llama_parity', 'inference_llama_parity_prefill', 'inference_llama_parity_decode', 'inference_llama_autopsy', 'inference_llama_dump_index', 'fd_gradients', 'train_parity_epochs_3', 'train_parity_epochs_5', 'train_runtime_parity_realistic', 'train_runtime_parity_stress', 'replay_determinism', 'profile_summary', 'perf_stat_summary', 'flamegraph_manifest', 'cachegrind_summary', 'asan_summary', 'vtune_summary', 'advisor_summary', 'likwid_summary', 'memory_signoff', 'perf_gate_report', 'embedding_dump', 'multimodal_bridge_report', 'full_network_graph', 'bridge_encoder_ir1', 'bridge_encoder_layout', 'bridge_encoder_call', 'bridge_encoder_lowered', 'xray_whisper_encoder', 'xray_ranking', 'xray_execution_trace', 'xray_execution_state', 'xray_decoder_pytorch', 'xray_qwen3vl_pytorch', 'xray_qwen3vl_llamacpp', 'xray_monotonic', 'analysis_checkpoints', 'training_logbook', 'training_checkpoint_history']\n Loaded 41 files\n\nGenerated: /home/runner/work/C-Kernel-Engine/C-Kernel-Engine/version/v8/.cache/training_workflow/qwen3_4layer_fp32/ir_report.html\n" } }, "performance": { "certification_loop_seconds": 30.833142042999953, "certification_workflow_tokens_per_second": 3243.263364484224, "export_compile_seconds": 7.231489522999709, "generated_c_optimizer_steps_per_second": 78.82498505843571, "generated_c_tokens_per_second": 20108.414555723397, "generated_c_training_seconds": 4.973042490390537, "generated_profile_ms": { "backward_ms": 2419.94384765625, "final_flush_ms": 0.147714999911841, "forward_ms": 2360.39990234375, "optimizer_ms": 41.76171875, "step_ms": 4972.894775390625 }, "optimizer_steps": 392, "peak_rss_bytes": 348270592, "setup_compile_seconds": 7.149311442999988, "tokens_per_second": 20108.414555723397, "tokens_per_second_scope": "generated_c_profiled_steps_plus_final_flush", "v8_inference_compile_seconds": 65.63879275799991 }, "failures": [], "summary_path": "version/v8/.cache/reports/training_workflow_latest.json", "note": "Current four-layer training-to-inference evidence." } }