[perf]Add fused Kimi image preprocessing (#47416)

Kevin_Xiong committed 4 months ago

[Perf] Cache `token_to_req_indices` for dsv4, 5x~6x kernel performance improvement (#47474)

Wentao Ye committed 4 months ago

[Bugfix][Spec Decode] Add missing draft_id_to_target_id to DSparkDeepseekV4ForCausalLM (#47429)

Laurent-Zhang committed 4 months ago

[Refactor] Remove multiple dead code (#47329)

Wentao Ye committed 4 months ago

[Bugfix][Distributed] Delegate MNNVL allreduce one-shot selection (#47589)

jesco committed 4 months ago

[Bugfix] Fix int32 overflow in triton_decode_attention page offsets (#47671)

Yifan Qiao committed 4 months ago

[CI] Skip test for checkpoint that was deleted (#47748)

Harry Mellor committed 4 months ago

[Bugfix][Core] Fix num_output_placeholders underflow with async scheduling + spec decode (#46066)

Ting SUN committed 4 months ago

[Doc] Clarify fastokens availability (#45813)

LiJzd committed 4 months ago

Revert "[Platform] Replace `torch.cuda.Event` with `torch.Event` (#47140)" (#47668)

Kunshang Ji committed 4 months ago

[Bugfix] [Gemma4] Fix Gemma4 MTP draft model layers ignoring quant_config (#47091)

Ayushman Singh committed 4 months ago

[Rust Frontend] Cache metric handles for scheduler & request stats (#47444)

Bugen Zhao committed 4 months ago

attention: pass None for unused args in unified attention TD path (#43597)

Artur Fierka committed 4 months ago

[CI] Use TTY for AMD CI tests for colored buildkite logs (#47730)

Nick Hill committed 4 months ago

[fix][run_batch]: respect proxy env vars when downloading media URLs (#47697)

Qiming Zhang committed 4 months ago

[ROCm][AITER] Directly Implement AITER Custom All-reduce in CudaCommunicator (#46065)

BadrBasowid committed 4 months ago

[Bugfix] Return HTTP 422 for unprocessable image URLs instead of 500 (#47165)

Alexis K. committed 4 months ago

fix(security): block request-level GPU video backend selection withou… (#47259)

Juan Pérez de Algaba committed 4 months ago

[bugfix] fix MOSS-Audio deepstack_input_embeds initialization in PP (#47607)

Yan Ma committed 4 months ago

[XPU][Bugfix] Do not transpose weight_scale_inv at load time (#47116)

Ma Jian committed 4 months ago