[Model] Honor cap_pixels_per_frame in Qwen3-VL memory profiling (#54380)

Devon Krisman committed about 2 months ago

[Bugfix] Reset cached Mamba align metadata on profiling teardown (#54044)

Canlin Guo committed about 2 months ago

[CI] Add explicit step keys to 18 hardware test steps (#54330)

Kevin H. Luu committed about 2 months ago

[Bugfix] Bound cache_salt length to prevent DoS via scheduler CPU exhaustion (#54353)

Juan Pérez de Algaba committed about 2 months ago

[Fix] Improve ROCm detection in WSL environments (#38434)

Yizhou committed about 2 months ago

[Rust Frontend][gRPC] Add audio and video media inputs (#53760)

Connor Carpenter committed about 2 months ago

[Bugfix][Multimodal] Release Qwen2.5-VL and Qwen3-VL RoPE caches with the model (#54346)

waizuichougou committed about 2 months ago

[ROCm][MLA] Reach FULL cudagraphs for AITER MLA speculative decoding (#51171)

Matt Qin committed about 2 months ago

[Bugfix][MoE] Enable cuBLAS out_dtype router GEMM on all CUDA archs (fixes family-120/GB10) (#54048)

hcl committed about 2 months ago

[CI][Ray] Fix flaky multi-node assignment test after placement-group teardown (#53621)

yzong-rh committed about 2 months ago

[CI][Test] Deflake test_mem.py sleep-mode asserts via allocator bookeeping (#54312)

Oxana Korzh committed about 2 months ago

[Bugfix][Model Runner V2][Spec Decode] Decouple the draft's gumbel noise stream from the target's (#54282)

Giancarlo Delfin committed about 2 months ago

[Bugfix][Frontend] Only echo the assistant turn in batched chat completions (#52529)

Kaif Kohari committed about 2 months ago

[Bugfix][Frontend] Validate stop_token_ids against vocab size (#54196)

jack committed about 2 months ago

[Perf] Reuse topk SparseMatrix routing metadata in GPT-OSS MoE forward (#45457)

Shenglei Fu committed about 2 months ago

[Nixl][PD] DCP support for MLA models (#50611)

Nicolò Lucchesi committed about 2 months ago

[Multimodal] Deprecate PyAV video decoder backend (#54231)

Isotr0py committed about 2 months ago

[Mypy] Fix typing for J models (#54130)

Taneem Ibrahim committed about 2 months ago

[ROCm][CI] Stage E gating (#50920)

Andreas Karatzas committed about 2 months ago

[Bugfix][Spec Decode] Capture the widest uniform decode batch by default (#50488)

Rahul Chalamala committed about 2 months ago