[Docs] Add Phi-3.5-mini-instruct to batch invariance tested models (#46396)

yuvalluria committed 3 months ago

[Spec Decode][DSpark] Add Gemma4-12B DSpark draft model (#47216)

Hangrui Cao committed 3 months ago

[Quant] Add `nvfp4_per_token` online MoE quantization (#48538)

Michael Goin committed 3 months ago

[Bugfix] Fix activation quantization dispatch for WNA4Int/WNA8Int (#48785)

HDCharles committed 3 months ago

[Model] Add Inkling MTP=1 support [3/N] (#48869)

Woosuk Kwon committed 3 months ago

[Bugfix][KV-transfer] MoRIIO: retry RDMA send-queue-full backpressure instead of failing the read (#47495)

limeward committed 3 months ago

[Helion] Fix degenerate scale_ub in kernel input generators (#48868)

Shangdi Yu committed 3 months ago

Fix Quark mxfp4 quantized model loading issue under mtp (#46757)

xiao-llm committed 3 months ago

[Model] Add Hopper FA4 relative attention for Inkling (#48858)

Woosuk Kwon committed 3 months ago

[Refactor] Move fla to third party (#48500)

Wentao Ye committed 3 months ago

[Model] Add PW CUDA graph support for Inkling [2/N] (#48822)

Woosuk Kwon committed 3 months ago

[KV Offload] Split tiering_lookup_delay into sync/async histograms (#47679)

Srinivas Krovvidi committed 3 months ago

[ROCm][CI] Avoid HIP init at config time via lazy aiter import in Quark OCP-MX (#48015)

music-dino committed 3 months ago

[Bug][Quantization] Fix humming is_layer_skipped for compressed-tensors "re:" ignore entries (#48507)

Zhongdongming Dai committed 3 months ago

[Perf][MoE] Write FlashInfer combine into final output (#47156)

Samuel Nordmann committed 3 months ago

[Frontend] Overlap preprocessing and computation for pooling models offline inference (#47699)

wang.yuqi committed 3 months ago

[CI/Build][Docker] Bump nvidia-cutlass-dsl to 4.6.0 and drop packaging workarounds (#47442)

Artem Perevedentsev committed 3 months ago

[ROCm][Perf][DSV4] Enable split sparse decode on gfx942 (#46275)

Tuukka Sarvi committed 3 months ago

[NIXL] Bump nixl to 1.3.1 (#47559)

ovidiusm committed 3 months ago

[KV Connector] Fix PD async scheduling race condition for hybrid attn models (#48481)

Artem Perevedentsev committed 3 months ago