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>
2 lines
281 B
JSON
2 lines
281 B
JSON
{"schema":1,"record_type":"footer_checkpoint","stream":"opprof-v1-dp0-pid2661256-1783869521184806299.jsonl","encoded_records":6687,"written_records":6687,"dropped_records":0,"last_step_index":6686,"checkpoint_wall_ns":1783869792959224277,"flush_interval_seconds":1.0,"final":true}
|