Gahow Wang 8e58b4033d Note dash1 lacks LLM gateway access (naive-completion deferred to dash0) ...

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

2026-06-17 09:55:39 +08:00

..

harness-vs-naive-20260616.md

Note dash1 lacks LLM gateway access (naive-completion deferred to dash0)

2026-06-17 09:55:39 +08:00

profile-driven-harness-design.md

Document profile-driven harness design

2026-05-12 21:09:29 +08:00

profile-driven-harness-implementation-20260512.md

Document 8-GPU harness ablation results for qwen27b and qwen235b prefill

2026-05-16 21:23:16 +08:00

qwen27b-chat-0-8k-ttft4s-tpot25-20260510.md

Document 8-GPU harness ablation results for qwen27b and qwen235b prefill

2026-05-16 21:23:16 +08:00

qwen27b-chat-0-8k-ttft4s-tpot25-gpu8-20260513.md

Document 8-GPU harness ablation results for qwen27b and qwen235b prefill

2026-05-16 21:23:16 +08:00

qwen27b-tp-sweep-20260616.md

Document 27B TP sweep: per-GPU rises sharply with TP (dense), opposite of MoE

2026-06-16 01:54:40 +08:00

qwen235b-thinking-prefill-ttft-3s6s9s-20260514.md

Document 8-GPU harness ablation results for qwen27b and qwen235b prefill

2026-05-16 21:23:16 +08:00

qwen235b-thinking-prefill-ttft-20260510.md

Document 8-GPU harness ablation results for qwen27b and qwen235b prefill

2026-05-16 21:23:16 +08:00

stop-a-validation-20260615.md

Record Stop-A boundary-guard A/B: correct verdict, ~38% replay saved

2026-06-15 16:57:53 +08:00

stop-b-e2e-27b-20260616.md

Document Stop-B end-to-end on dense 27B: the improving climb + no-regression

2026-06-16 18:07:00 +08:00

stop-b-e2e-20260615.md

Flag Stop-B e2e per-GPU trajectory as non-benchmark (saturation + smoke regime)

2026-06-15 18:40:38 +08:00