Bugfix: Pass router logits dtype in nemotron shared experts (#32669)

amirkl94 committed 9 months ago

Make `mypy` opt-out instead of opt-in (#33205)

Harry Mellor committed 9 months ago

[Misc] Remove missed `pad_for_cudagraph` (#33283)

Lucas Wilkinson committed 9 months ago

[Doc] Update outdated link to Ray documentation (#32660)

graftim committed 9 months ago

Adding optional speculator tests for larger models (#32943)

shanjiaz committed 9 months ago

[PluggableLayer][2/N] Apply PluggableLayer to linear layers (#33152)

whx committed 9 months ago

support returning tokenids in responses api (#33212)

cmunley1 committed 9 months ago

[BugFix] Fix EPLB fail for MoeFP4 model with Marlin backend (#33262)

Ilya Markov committed 9 months ago

[Doc]: fixing multiple typos in diverse files (#33256)

Didier Durand committed 9 months ago

[Bugfix] Fix Qwen3-VL-Reranker load. (#33298)

wang.yuqi committed 9 months ago

[CI/Build][BugFix] fix cuda/compat loading order issue in docker build (#33116)

Pengchao Wang committed 9 months ago

[Release] [ROCm] Remove old build step (#33316)

TJian committed 9 months ago

[Misc][Build] Lazy load cv2 in nemotron_parse.py (#33189)

Kiersten Stokes committed 9 months ago

[Release] [CI] Optim release pipeline (#33156)

TJian committed 9 months ago

Fix tool call indexing double-counting (#33141)

wangln19 committed 9 months ago

[Refactor] Define MM data parser in processing info instead of processor itself (#33260)

Cyrus Leung committed 9 months ago

[ez] Delete more torch version checks <= 2.8 (#33288)

Angela Yi committed 9 months ago

[Misc] Add orozery to CODEOWNERS (core, kv_transfer, kv_offload) (#33227)

Or Ozeri committed 9 months ago

[Bugfix] Register fp8 cutlass_group_gemm as supported for only SM90+SM100 (#33285)

Michael Goin committed 9 months ago

[UX] Remove noisy CT UnquantizedLinearMethod warn (#33273)

Michael Goin committed 9 months ago