The packaging commit shipped the code and a summary of the numbers. It left out
everything the numbers are measured *against*, which is most of what makes them
checkable.
Added:
baselines/ llama.cpp runs (thread sweeps, NUMA sweeps, the
interleaved A/B passes behind the headline decode
figure) and the quality references: each build
against its own bf16 over 12 greedy prompts.
benchmarks/.../results/ 102 artifacts backing the published figures, each
with a provenance block naming the model,
quantization, thread set and build SHAs.
docs/engines/ the engine analyses -- llama.cpp, ExecuTorch, MNN,
ncnn, vLLM, vLLM-Omni -- and the cross-engine and
device-simulation comparisons.
docs/analysis/ the full study documents the summaries distil.
docs/edge/comparisons.md where this fork sits against llama.cpp and why:
decode is llama.cpp's, prefill is ours by ~2x, and
since neither engine exceeds 39% of the measured
222 GB/s ceiling the decode gap is the execution
model rather than the kernels.
docs/edge/environment.md pins and machine, and the correction below.
On "the modified vllm": there isn't one, and the doc says so plainly because it
is the obvious assumption. The vLLM checkout is clean upstream (0 commits
ahead, 0 uncommitted) and the installed wheel is stock 0.28.0+cpu. This fork
changes vLLM's CPU *behaviour* through vLLM's own extension points at import
time -- three CPU patches in vllm_omni/patch.py, the quantization-config
registry, the IR op registry, platform hooks -- specifically so the changes
survive a wheel upgrade rather than being lost on it. One of those patches is
why "KV sizing is not a memory lever" looked true for months: CpuPlatform
overwrote kv_cache_memory_bytes unconditionally, so the setting had no effect.
There is no patch series to apply; pin the version and let patch.py do the rest.
.gitignore gains explicit exceptions: upstream's blanket `*.json` and `results/`
rules were silently eating the measurement artifacts, which in this fork are the
substance rather than build output. Binary captures stay ignored.
No vLLM or llama.cpp source is redistributed -- only output produced with them.
NOTICE records both.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L1ax5wYLA6XbqEL1i8L9gZ
1 line
8.5 KiB
JSON
1 line
8.5 KiB
JSON
{"execution_summary": {"estimated_inference_time": 838, "estimated_inference_peak_memory": 138752000, "first_load_time": 104984, "first_load_peak_memory": 113438720, "warm_load_time": 104086, "warm_load_peak_memory": 125595648, "compile_time": 0, "compile_peak_memory": 0, "compile_memory_increase_range": [0, 0], "compile_memory_peak_range": [0, 0], "first_load_memory_increase_range": [0, 8178736], "first_load_memory_peak_range": [7778304, 16937392], "warm_load_memory_increase_range": [0, 1753392], "warm_load_memory_peak_range": [872448, 11059632], "inference_memory_increase_range": [0, 1069104], "inference_memory_peak_range": [0, 9035184], "all_compile_times": [], "all_first_load_times": [104984], "all_warm_load_times": [104086], "all_inference_times": [1563, 1071, 986, 987, 979, 919, 891, 900, 880, 880, 874, 1063, 935, 943, 935, 941, 937, 952, 846, 867, 840, 850, 887, 883, 904, 901, 867, 868, 897, 899, 867, 867, 867, 854, 890, 953, 928, 956, 1122, 997, 1083, 1165, 908, 911, 917, 1029, 1172, 913, 953, 929, 933, 943, 960, 947, 876, 894, 953, 979, 917, 932, 917, 916, 942, 932, 962, 922, 924, 907, 862, 917, 928, 930, 933, 940, 942, 907, 930, 936, 916, 838, 845, 866, 885, 871, 873, 873, 887, 877, 966, 890, 869, 874, 862, 871, 856, 842, 872, 873, 867, 877]}, "execution_detail": [{"name": "Input", "type": "Input", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 478820}, {"name": "rms_norm_/layers.0/Mul_1_output_0", "type": "rms", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 53145}, {"name": "OpId_20", "type": "", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 126}, {"name": "/layers.0/q_k_v_proj/MatMul", "type": "/layers.0/q", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/q_k_v_proj/MatMul_post_reshape", "type": "/layers.0/q", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "node_Slice_497", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 83913}, {"name": "node_Slice_500", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 9103}, {"name": "node_Slice_503", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 16251}, {"name": "/layers.0/Reshape", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Reshape_1", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Reshape_2", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/g_proj/MatMul", "type": "/layers.0/g", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 5974}, {"name": "/layers.0/g_proj/MatMul_post_reshape", "type": "/layers.0/g", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Sigmoid", "type": "/layers.0/Sigmoid", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 2653}, {"name": "/layers.0/Reshape_3", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1451}, {"name": "node_Slice_393", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 328}, {"name": "node_Slice_396", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 355}, {"name": "/layers.0/Mul_2", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 10257}, {"name": "/layers.0/Neg", "type": "/layers.0/Neg", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 492}, {"name": "/layers.0/Concat", "type": "/layers.0/Concat", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 865}, {"name": "/layers.0/Mul_3", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 12237}, {"name": "/layers.0/Add_1", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1279}, {"name": "node_Slice_442", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 5166}, {"name": "node_Slice_445", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 5196}, {"name": "/layers.0/Mul_4", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 11283}, {"name": "/layers.0/Neg_1", "type": "/layers.0/Neg", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1134}, {"name": "/layers.0/Concat_1", "type": "/layers.0/Concat", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1710}, {"name": "/layers.0/Mul_5", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 11689}, {"name": "/layers.0/Add_2", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 2282}, {"name": "/layers.0/Reshape_4", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 198}, {"name": "/layers.0/Concat_2", "type": "/layers.0/Concat", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 20515}, {"name": "/layers.0/Concat_3", "type": "/layers.0/Concat", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 17531}, {"name": "/layers.0/Concat_3_output_0_converted_unsigned_symmetric", "type": "/layers.0/Concat", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 128046}, {"name": "/layers.0/Transpose", "type": "/layers.0/Transpose", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Transpose_output_0_converted_unsigned_symmetric", "type": "/layers.0/Transpose", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/MatMul", "type": "/layers.0/MatMul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 367695}, {"name": "/layers.0/Mul_6", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1231}, {"name": "/layers.0/Add_3", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 14654}, {"name": "/layers.0/Softmax", "type": "/layers.0/Softmax", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 11444}, {"name": "/layers.0/MatMul_1", "type": "/layers.0/MatMul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 57445}, {"name": "/layers.0/Reshape_5", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Mul_7", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 24919}, {"name": "/layers.0/out_proj/MatMul", "type": "/layers.0/out", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 52505}, {"name": "/layers.0/out_proj/MatMul_post_reshape", "type": "/layers.0/out", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Add_4", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 10812}, {"name": "rms_norm_/layers.0/Mul_9_output_0", "type": "rms", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 9451}, {"name": "/layers.0/gate_proj/MatMul", "type": "/layers.0/gate", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/gate_proj/MatMul_post_reshape", "type": "/layers.0/gate", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "node_Gelu_0", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 436131}, {"name": "/layers.0/up_proj/MatMul", "type": "/layers.0/up", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 403371}, {"name": "/layers.0/up_proj/MatMul_post_reshape", "type": "/layers.0/up", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Mul_12", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 23019}, {"name": "/layers.0/down_proj/MatMul", "type": "/layers.0/down", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 405361}, {"name": "/layers.0/down_proj/MatMul_post_reshape", "type": "/layers.0/down", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Add_7", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 5781}, {"name": "/layers.0/Add_2_identity", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Reshape_2_identity", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "Output", "type": "Output", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 4086}]} |