[Bugfix][Kernel] fix merge attn states when both prefix and suffix are empty (#28181)

courage17340 committed about 1 year ago

[V0 deprecation]clean up is_v1_supported_oracle (#28116)

wangxiyuan committed about 1 year ago

[CPU] Enable torch profiling (#28130)

Aditya Tewari committed about 1 year ago

Fix hard-coded parameter name in gemma3n.py (#27946)

Seungduk Kim committed about 1 year ago

Patch Mistral Tokenizer (#28146)

Julien Denize committed about 1 year ago

[flashinfer] fix FI all2all with FI cutlass moe (#28166)

Xiaozhu Meng committed about 1 year ago

[CI] Add compile/test_multimodal_compile.py to CI (#28151)

gmagogsfm committed about 1 year ago

Make the cv2 dependency optional (#27780)

Jacob Zhong committed about 1 year ago

[Core][MM] Use non-blocking CPU-GPU copy of multimodal data (#28141)

Lukas Geiger committed about 1 year ago

[Chore] Clean up deepseek v2/v3 config copy (#28055)

Isotr0py committed about 1 year ago

[Misc] Remove the duplicate code (#28111)

Chauncey committed about 1 year ago

[CI Failure] `nm-testing/Qwen2-0.5B-Instruct-FP8-SkipQKV` was removed from HF. Skip it in tests (#28170)

Vadim Gimpelson committed about 1 year ago

[Feature] Enable TP + EP `shared_experts` overlap with router, 3.7% E2E performance improvement (#28164)

Wentao Ye committed about 1 year ago

[Bug] Fix env string `"0"` same to `True` (#28159)

Wentao Ye committed about 1 year ago

[Bug] Fix cpu disable shared_experts `VLLM_DISABLE_SHARED_EXPERTS_STREAM` (#28157)

Wentao Ye committed about 1 year ago

[PERF] Decouple projections from GDN custom op. Attempt 2 (#28083)

Vadim Gimpelson committed about 1 year ago

[Debugging] Add annotation for easier trace analysis (#22496)

Dayeol Lee committed about 1 year ago

[Core][Hybrid allocator + connector 2/n] Unify `remove_skipped_blocks` by `get_last_useful_token` (#25431)

Kuntai Du committed about 1 year ago

[CI/Build] Fix `test_defaults_with_usage_context` in AMD CI (#27926)

Zhewen Li committed about 1 year ago

[CI/Build] Update checking logic in cutlass_group_gemm_supported (#27948)

Zhewen Li committed about 1 year ago