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": 0.016591107009,
|
|
"cell": "tp4_mns8",
|
|
"early_stop_reason": "",
|
|
"early_stopped": false,
|
|
"exact_output_count": 16,
|
|
"feasible": false,
|
|
"interval": {
|
|
"elapsed_s": 6.775592838,
|
|
"end_mono_ns": 201547098885716,
|
|
"end_wall_ns": 1783869974225312732,
|
|
"start_mono_ns": 201540323292878,
|
|
"start_wall_ns": 1783869967449719953
|
|
},
|
|
"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": "99f2dcf01a54ab81776395fdef964b2a109d37fd8d44f971f3723e99a012b35e",
|
|
"arrival_s": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 3.0116999999999736,
|
|
"min": 0.0,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
},
|
|
"count": 16,
|
|
"long_gt4096": 5,
|
|
"offered_req_s": 0.26666666666666666,
|
|
"offered_req_s_per_gpu": 0.06666666666666667,
|
|
"raw_input_tokens": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 8150.0,
|
|
"min": 810.0,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
},
|
|
"raw_length_order_sha256": "3b3e3258b82a4168ea8c3836458779878ca106893e1af47b23576e89de4b245f",
|
|
"request_id_order_sha256": "e321d81e19450f59c2ac022f6ae487ec4c72024453d015376c2df60b97167d76"
|
|
},
|
|
"slo_pass_count": 10,
|
|
"study_sha256": "76e9d642ba9570917cfeeea848ed92f27904e87760eedabd48805b84e82eca26",
|
|
"tp": 4,
|
|
"tpot_ms": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 18.231644558923787,
|
|
"min": 5.76333962215888,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
},
|
|
"ttft_ms": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 3136.144638003316,
|
|
"min": 1031.9823420140892,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
}
|
|
}
|