Gahow Wang
d5b276180d
Workload-conditioned operator profiling on patched vLLM 0.24.0 + Qwen3-30B-A3B/H20. H1b PASS (irregular patterns carry +23-45pp R64 raggedness, 8-45% token-efficiency loss vs rectangular controls); mechanism decomposition kills the padding narrative and finds the arrival-uniformization artifact (-12.9%); cross-version churn surface shows TP2/MNS64 -29.4% across vLLM 0.20->0.24 while the argmax held. Raw Layer-1 JSONL streams (507 MB) stay on disk, git-ignored; footer sidecars and metrics are tracked. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
75 lines
1.9 KiB
JSON
75 lines
1.9 KiB
JSON
{
|
|
"anchor": 0.49609375,
|
|
"cell": "tp2_mns8",
|
|
"early_stop_reason": "",
|
|
"early_stopped": false,
|
|
"exact_output_count": 16,
|
|
"feasible": false,
|
|
"interval": {
|
|
"elapsed_s": 9.026111125,
|
|
"end_mono_ns": 201115775651265,
|
|
"end_wall_ns": 1783869542902078194,
|
|
"start_mono_ns": 201106749540140,
|
|
"start_wall_ns": 1783869533875967269
|
|
},
|
|
"invariants": {
|
|
"arrival_nondecreasing": true,
|
|
"exact_output_or_failed": true,
|
|
"outcomes_cover_selected": true,
|
|
"raw_lengths_present": true,
|
|
"selected_nonempty": true,
|
|
"warmup_16": true,
|
|
"warmup_exact_16": true,
|
|
"warmup_long": true
|
|
},
|
|
"kind": "warmup",
|
|
"mns": 8,
|
|
"observed_count": 16,
|
|
"pass_rate": 0.625,
|
|
"schema": 1,
|
|
"selection": {
|
|
"arrival_order_sha256": "818d316f68acc6502ed33a877bf88c2ba9b57a1ff6d6780621db0d30eddf3494",
|
|
"arrival_s": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 3.4938000000000105,
|
|
"min": 0.0,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
},
|
|
"count": 16,
|
|
"long_gt4096": 7,
|
|
"offered_req_s": 0.26666666666666666,
|
|
"offered_req_s_per_gpu": 0.13333333333333333,
|
|
"raw_input_tokens": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 7270.0,
|
|
"min": 71.0,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
},
|
|
"raw_length_order_sha256": "2eb3b78b31efc9354c64c8caa95f08353330553d1ab745a8e0ceab038a3a3d5e",
|
|
"request_id_order_sha256": "0009512d07ec4c5bb2dd1283c767967a6dfd5bc4c7b7d65fb26169d39a98832e"
|
|
},
|
|
"slo_pass_count": 10,
|
|
"study_sha256": "9474f0d0b53579f1db852ca68abfb0b96ba43ae4e17738118bf8e3209eb09ece",
|
|
"tp": 2,
|
|
"tpot_ms": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 39.46136329144306,
|
|
"min": 7.347993740140952,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
},
|
|
"ttft_ms": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 4594.941660994664,
|
|
"min": 762.8166690119542,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
}
|
|
}
|