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 9, 13:33 UTC
main
[Frontend] Reduce mixin usage in serving pooling (#33101)
Cyrus Leung committed 10 months ago
[Perf] avoid duplicate mem_get_info() call in get_current_memory_usage (#33064)
Paco Xu committed 10 months ago
Vincent Gimenes committed 10 months ago
Fix IndexError with encoder-decoder models when using Custom Paged Attention (#33112)
Strahinja Stamenkovic committed 10 months ago
fix: preserve native tool call ID in multi-turn tool calling (#32768)
wangln19 committed 10 months ago
[MoE Refactor] Integrate Naive Prepare Finalize into MK (#32567)
Robert Shaw committed 10 months ago
[Model Runner V2] Remove UvaBufferPool for cpu->gpu copy (#33055)
Woosuk Kwon committed 10 months ago
[Bugfix][TPU] Return a Default fp8 MoE Backend (#32908)
XiongfeiWei committed 10 months ago
[Bugfix][MXFP4] Call `trtllm_fp4_block_scale_moe` with kwargs (#33104)
Pengchao Wang committed 10 months ago
[fix] CPUDNNLGEMMHandler pointer baked into inductor artifact (#32913)
dolpm committed 10 months ago
[Logging] add `--disable-access-log-for-endpoints` CLI option (#30011)
Jared Wen committed 10 months ago
[Refactor] Remove unused `_moe_permute` function (#33108)
Wentao Ye committed 10 months ago
[ci] Sync test areas with test-pipeline.yaml and enable new pipeline generator (#33080)
Kevin H. Luu committed 10 months ago
[Bugfix] Fix Dtypes for Pynccl Wrapper (#33030)
Robert Shaw committed 10 months ago
[Model] Bump transformers version for test registry (#33100)
Cyrus Leung committed 10 months ago
[Bugfix] Fix Voxtral streaming slot_mapping (#33073)
Nicolò Lucchesi committed 10 months ago
[FIX] Always support TP > 4 for FP4 Gemm (#31099)
danielafrimi committed 10 months ago
Remove unused logic in `models/mistral.py` (#33095)
Andy Lo committed 10 months ago
[CI] Fix AssertionError: MCP tool call not found in output_messages (#33093)
Chauncey committed 10 months ago
[ROCm][Bugfix] Fix ptpc scale load issue for fused shared expert path in deepseek mtp (#33018)
Pleaplusone committed 10 months ago