[feature]Add int4 quantization support for emulation moe backend (#48451)

qli88 committed 3 months ago

[Attention] Mirror Triton KV dtype checks in MLA (#47060)

Mike G committed 3 months ago

[Compilation] Skip x.size(dim) in _decompose_size_nodes (#42543)

nemanjaudovic committed 3 months ago

[Tests] Gate Step3VL under Transformers v5 (#44349)

BRIJ RAJ KISHORE committed 3 months ago

[Refactor] Move iteration logging to the frontend (#46647)

Max Hu committed 3 months ago

[XPU] Add DSpark speculative decoding support for DeepSeek-V4 (#47677)

Ma Jian committed 3 months ago

[ROCm][Perf] Optimize sparse attention prefill kernel for DeepSeek-V4 (#48519)

kliuae committed 3 months ago

[Bugfix] Fix local speculators with dots in the name from classifying as custom_class (#48754)

Michael Goin committed 3 months ago

[BugFix][MLA] Support kv_cache_dtype_skip_layers for MLA attention (#47309)

ruikangliu committed 3 months ago

[Perf] Optimize `fused_topk_bias` for DSv4, 1.5~2x kernel performance improvement (#47463)

Wentao Ye committed 3 months ago

Add giuseppegrossi to rocm label auto cc action (#48643)

Giuseppe Grossi committed 3 months ago

[Feature] Migrate moe sp support to non-torch compiled path for GLM5.2 (#47881)

Wentao Ye committed 3 months ago

[ROCm][Perf] DSv4 two-stage compressor kernel for HCA prefill (#47718)

kliuae committed 3 months ago

[Bugfix][NVFP4 MoE] Pad gated intermediate to 64 for FlashInfer TRT-LLM shuffle (M%128) (#46880)

Mike G committed 3 months ago

[CI][PD] Add optional/nightly DSv4 Disaggregated eval (#42310)

Nicolò Lucchesi committed 3 months ago

[Bugfix] Fix FlashInfer non-causal draft attention (DFlash/DSpark) on Blackwell (#48167)

Michael Goin committed 3 months ago

[CI][ROCm] Retry failed Docker build steps once (#48773)

Micah Williamson committed 3 months ago

[Bugfix] Initialize draft CUDA-graph keys for the native draft_model proposer (#47460)

avalliappan-nvidia committed 3 months ago

[Docs] Document pooling config resolution (#48497)

Taneem Ibrahim committed 3 months ago

[Fix] Align OpenAI vllm_xargs value types across request schemas (#48252)

Sage committed 3 months ago