A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 2, 08:30 UTC
main
[Refactor] Remove dead current_tool_name_sent assignments from tool parsers (#43997)
Flora Feng committed 5 months ago
offload prompt_embeds decode in render_prompts_async to avoid blocking (#43792)
Gagan Dhakrey committed 5 months ago
[PERF]MiniMax-M2 gate kernel (#38445)
Jee Jee Li committed 5 months ago
Revert "[MoE Refactor] Migrate MoeWNA16Method quantization to MK orac… (#44033)
bnellnm committed 5 months ago
[CI] Fix smoke test step key to bypass block gate (#43974)
Kevin H. Luu committed 5 months ago
[CI] Remove duplicate Harmony test coverage (#44023)
Flora Feng committed 5 months ago
[MoE Refactor] Remove supports_expert_map (#43108)
bnellnm committed 5 months ago
[MoE Refactor] Migrate MoeWNA16Method quantization to MK oracle (#42647)
bnellnm committed 5 months ago
[Fronten] Clean up stop_token_ids override for Harmony (#44009)
yzong-rh committed 5 months ago
[Metrics] Exclude KV transfer tokens from iteration_tokens_total (#43346)
Tyler Michael Smith committed 5 months ago
[Feature] SSL support for dp supervisor (#43688)
Wentao Ye committed 5 months ago
Add @khluu to CODEOWNERS (#44019)
Kevin H. Luu committed 5 months ago
[CI] Remove redundant test_chat_with_tool_reasoning.py (#44011)
Flora Feng committed 5 months ago
[CI] Make Model Executor test hangs fail fast with a traceback (#43971)
Kevin H. Luu committed 5 months ago
[Bug] Fix torch device issue for MOE permute (#44005)
Wentao Ye committed 5 months ago
[Bugfix] Fix Ray placement group allocation with grouped nodes (#43998)
czhu-cohere committed 5 months ago
[Bugfix] Use storage_block_size in KV cache reshape for compressed specs (DeepSeek V4) (#43988)
qizixi committed 5 months ago
[EPLB] Make async EPLB default (#43219)
Ilya Markov committed 5 months ago
[MoE Refactor] WNA16 MoE backend selection into oracle module (#42553)
bnellnm committed 5 months ago
[Bugfix] Disable allreduce_rms_fusion when pipeline_parallel_size > 1 (#43616)
qizixi committed 5 months ago