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": 0.75,
|
|
"cell": "tp2_mns64",
|
|
"early_stop_reason": "",
|
|
"early_stopped": false,
|
|
"exact_output_count": 16,
|
|
"feasible": false,
|
|
"interval": {
|
|
"elapsed_s": 7.415686836,
|
|
"end_mono_ns": 201114188199246,
|
|
"end_wall_ns": 1783869541314626167,
|
|
"start_mono_ns": 201106772512410,
|
|
"start_wall_ns": 1783869533898939505
|
|
},
|
|
"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": 64,
|
|
"observed_count": 16,
|
|
"pass_rate": 0.75,
|
|
"schema": 1,
|
|
"selection": {
|
|
"arrival_order_sha256": "97f6ded38afa2a0fa9729d0553429b3df62e3d4e99710f4c19a706b1eba8fdaa",
|
|
"arrival_s": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 2.723000000000002,
|
|
"min": 0.0,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
},
|
|
"count": 16,
|
|
"long_gt4096": 5,
|
|
"offered_req_s": 0.26666666666666666,
|
|
"offered_req_s_per_gpu": 0.13333333333333333,
|
|
"raw_input_tokens": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 7233.0,
|
|
"min": 71.0,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
},
|
|
"raw_length_order_sha256": "76f4ce7bc309a1108d5d4b24ba24b374c14483a3dd35cd5648fbeafca60221ba",
|
|
"request_id_order_sha256": "b38587f14f8c188d131b4c5c8411951ec571999fb2ac6be6517f1ad952208852"
|
|
},
|
|
"slo_pass_count": 12,
|
|
"study_sha256": "9474f0d0b53579f1db852ca68abfb0b96ba43ae4e17738118bf8e3209eb09ece",
|
|
"tp": 2,
|
|
"tpot_ms": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 48.44081654317868,
|
|
"min": 9.34166979529587,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
},
|
|
"ttft_ms": {
|
|
"distinct_n": 16,
|
|
"finite_n": 16,
|
|
"max": 3932.37450599554,
|
|
"min": 783.6844140256289,
|
|
"missing_n": 0,
|
|
"n": 16
|
|
}
|
|
}
|