Repository navigation

perf(qwen27 cuda): reduce repeated-prompt TTFT - #747

Closed

jrb-ai wants to merge 24 commits into

ashhart:mainfrom

jrb-ai:perf/qwen27-pr726-ttft

Closed

perf(qwen27 cuda): reduce repeated-prompt TTFT#747
jrb-ai wants to merge 24 commits into
ashhart:mainfrom
jrb-ai:perf/qwen27-pr726-ttft

Commits on Oct 10, 2026
Commits on Oct 11, 2026
- Native image input: images from a request's image parts, prepared before admission; GLM-5.3-Flash reads them on Metal
  BobClawblaw
  committed
  907e013
- Native image input: --vision-urls fetches public HTTPS image URLs as 0.6.6 did
  BobClawblaw
  committed
  47c4098
Commits on Oct 12, 2026
Commits on Oct 13, 2026
- Merge main (1.0.9) into qwen27-cuda-1.0.7
  BobClawblaw
  committed
  e3b7210
- test(qwen27): integrate current native CUDA with PR726 1.0.9
  jrb-ai
  committed
  c7c77bc
- test(qwen27): apply PR747 TTFT increment to current native backend
  jrb-ai
  committed
  6b07f77
- test(qwen27): reuse quantized KV host rows during drafter startup
  jrb-ai
  committed
  a356e98
- fix(cuda): remove duplicate integrated fatbin exports
  jrb-ai
  committed
  5b6a80c
- test(qwen27): port qualified post-first capture to current native
  jrb-ai
  committed
  e22ab6d
Commits on Oct 14, 2026
- Sync shared CUDA operators to upstream bf3f16a2
  jrb-ai
  committed
  5cbd313
- perf(qwen27 cuda): publish validated native TTFT candidate
  jrb-ai
  committed
  841b77c