[Bugfix][SM120][MLA] Disable dense prefill for FlashInfer sparse MLA (#51395)

Tommy Asai committed 2 months ago

[Bugfix] Add forward_xpu to XDRotaryEmbedding for HunyuanOCR on XPU (#52174)

Jakub Byczkowski committed 2 months ago

[ROCm][CI] add Aiter ops tests (#52208)

Divakar Verma committed 2 months ago

[Perf][Structured Output] Skip unused request-local reasoners (#52573)

Bugen Zhao committed 2 months ago

[CI][AMD] Improve Kubernetes failure diagnostics (#52264)

Andreas Karatzas committed 2 months ago

[ModelRunnerV2] Support prompt embeds (#42963)

Canlin Guo committed 2 months ago

[ROCm][AMD][Installation] add LMCache kv-connector installation and runtime packages to docker image (#51208)

Hongxia Yang committed 2 months ago

[ROCm][CI] Expand AITER W4A4 MoE Coverage (#52647)

Micah Williamson committed 2 months ago

[Rust Frontend][gRPC] Advertise LoRA capabilities (#52031)

Connor Carpenter committed 2 months ago

[Spec decode] Support Kimi-K3 DCP with DSpark (#52188)

Wei Zhao committed 2 months ago

Relax CuPy constraint to only exclude 14.1.0 (#44284)

Kevin H. Luu committed 2 months ago

[Bugfix] Account for local DP workers in startup thread allocation (#52385)

crZhao committed 2 months ago

[ROCm][CI] Restore Torch defaults and type DSV4 scratch buffers (#52566)

Andreas Karatzas committed 2 months ago

[Perf][MoE] Optimize deepep_v2 receiver CPU Overhead (#51114)

Lucas Wilkinson committed 2 months ago

[Bugfix][CI] Release the shared ColBERT engine before `test_colbert_hf_comparison` (#52608)

stefankoncarevic committed 2 months ago

fix: prevent PyNvVideoCodec decoder slot limit bypass via ClassVar shadowing (#52126)

Juan Pérez de Algaba committed 2 months ago

[BugFix] lora_base_layer / routed_experts order in expert param mapping (#52552)

ℍ𝕠𝕝𝕝𝕠𝕨 𝕄𝕒𝕟 committed 2 months ago

[3/N][Feat][Perf] Add new warmup infrastructure for JITs. Add provider registry and orchestration for JIT warmup (#50174)

Roberto L. Castro committed 2 months ago

Support DSpark configs with `architectures=DSparkDraftModel` + `model_type=qwen3` (#52197)

Misha Goin committed 2 months ago

[Refactor] Simplify B12X linear kernels and warmup (#52368)

Misha Goin committed 2 months ago