Commit Graph

470 Commits

Author SHA1 Message Date
10567da523 Freeze Qwen235 ablation operator profiles 2026-07-19 17:11:26 +08:00
4ca295af0b Match FlashInfer TP8 workspace initialization 2026-07-19 16:54:24 +08:00
80ab724608 Run Qwen235 profiling from clean remote worktree 2026-07-19 16:47:12 +08:00
506e63a633 Add measured collective profile gate for Qwen235 2026-07-19 16:45:35 +08:00
f4a75aa8e4 Update prefix trace parser test contract 2026-07-19 15:35:17 +08:00
9c34650746 Merge simulator fidelity work into feature/sim 2026-07-19 15:34:28 +08:00
d17a2d4ab0 Merge fixed-PD pressure probe into feature/sim 2026-07-19 15:34:23 +08:00
3d3878c5aa Summarize Frontier selection regret 2026-07-19 15:31:16 +08:00
4c8d581a5b Track simulator fidelity experiment artifacts 2026-07-19 15:31:09 +08:00
e0ea7e9961 Support scp GPU fleet sync 2026-07-19 15:30:54 +08:00
fbf0f7c50b Set Qwen235 profile KV cache layout 2026-07-19 11:47:36 +08:00
81f3a5c76d Profile Qwen235 true mixed attention 2026-07-19 11:45:08 +08:00
71dbf80ed4 Match Qwen235 Frontier chunked prefill 2026-07-19 11:35:27 +08:00
5f48f7ec8b Avoid stale exact-trace HTTP connections 2026-07-19 10:58:38 +08:00
42e4ae3422 Avoid occupied Qwen235 trace ports 2026-07-19 08:10:29 +08:00
b72de0fd15 Wait for Qwen235 GPU memory cleanup 2026-07-19 07:07:08 +08:00
b2f97927de Expose nvcc to Qwen235 serving workers 2026-07-19 06:42:28 +08:00
979f179a47 Fix Qwen235 real surface output expansion 2026-07-19 06:29:47 +08:00
dfd9646d2c Freeze Qwen235 KV cache layout for profiling 2026-07-19 06:24:02 +08:00
922d66c5c1 Align Qwen235 attention profile batch limit 2026-07-19 06:21:33 +08:00
ac5061fa5d Align Qwen235 attention profile sequence limit 2026-07-19 06:19:08 +08:00
d563c30b42 Set Qwen3 model type for vLLM profile dispatch 2026-07-19 06:12:11 +08:00
a21382c8aa Raise Q235 profiler token limit 2026-07-19 05:45:37 +08:00
5067bc2cb1 Expose nvcc to Q235 FlashInfer profiling 2026-07-19 05:42:26 +08:00
39e4719b28 Run Qwen235 profilers from Frontier source 2026-07-19 03:00:04 +08:00
79e9870975 Defer Qwen235 campaign until Qwen30 completion 2026-07-19 02:54:32 +08:00
80a067e3a5 Orchestrate Qwen235 four-case fidelity campaign 2026-07-19 02:52:54 +08:00
3e32ea609f Cover Qwen235 MNS128 graph buckets 2026-07-19 02:48:51 +08:00
40bac6dbf4 Add Qwen235 same-stack Frontier profiling campaign 2026-07-19 02:46:58 +08:00
e631f6a269 Add Qwen235 Frontier latency surface runner 2026-07-19 02:44:55 +08:00
a3c8cb5808 Add Qwen235 Frontier MoE portability smoke 2026-07-19 02:42:28 +08:00
f4813cf537 Add Qwen235 vLLM 0.20 real surface runner 2026-07-19 02:39:16 +08:00
c8383c9c4c Launch Qwen30 fixed pressure comparison 2026-07-19 02:25:15 +08:00
7ea9635878 Record fixed PD pressure calibration result 2026-07-19 01:45:17 +08:00
33b73afe9b Narrow fixed PD pressure sweep around load knee 2026-07-19 01:26:09 +08:00
909a80f0a6 Reuse validated FlashInfer cache for pressure probe 2026-07-19 01:13:50 +08:00
9837fa5133 Match Qwen30 probe file descriptor limit 2026-07-19 01:07:57 +08:00
6726318792 Profile fixed PD workload pressure against trace anchor 2026-07-19 00:54:48 +08:00
e98608911e Resume valid Qwen30 latency cells safely 2026-07-18 17:27:12 +08:00
f11daa6776 Fix Qwen30 exact trace served-model routing 2026-07-18 11:56:43 +08:00
b9523cef5c Pin DeepGEMM nvcc for Qwen235 smoke 2026-07-18 11:15:22 +08:00
b7f9cef9c5 Keep Qwen30 JIT cache scoped to experiment 2026-07-18 11:11:10 +08:00
ecb45cf762 Extend Qwen30 JIT readiness for latency surface 2026-07-18 11:09:22 +08:00
686b050517 Isolate Qwen235 DeepGEMM JIT cache 2026-07-18 10:57:48 +08:00
28ffb22bee Create Qwen30 case output before launch logging 2026-07-18 10:50:11 +08:00
a01c1206ab Launch frozen Qwen30 latency case surfaces 2026-07-18 10:48:45 +08:00
5fc0b48fa1 Extend Qwen235 smoke readiness deadline 2026-07-18 10:32:21 +08:00
7a437b4d91 Allow fixed runtime preflight resume 2026-07-18 01:11:37 +08:00
12705411f9 Record Qwen235 portability experiment gate 2026-07-18 01:08:54 +08:00
ddacc5f7a6 Add Qwen235 vLLM 0.20 compatibility gate 2026-07-18 01:08:08 +08:00