vllm
https://github.com/vllm-project/vllm
Python
A high-throughput and memory-efficient inference and serving engine for LLMs
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
Python not yet supported37 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- buffer size insuffient Dspark sd for FlashInfer MNNVL allreduce
- [Feature][Core] Allow expert override for insufficient KV cache capacity
- [Bugfix] Skip MoE padding across router paths
- [Bug][P/D]: NIXL handshake failure with asymmetric TP in PD disaggregation (pTP2 + dTP4)
- [Bug]: override-generation-config silently ignores presence_penalty / frequency_penalty (missing from available_params whitelist)
- perf: Use fast driver P2P check on compute capability 12.x
- [Bug]: --load_format fastsafetensors produces corrupted/incoherent generation for DeepSeek-V4-Flash-0731 with TP+EP (works with default loader)
- [Bug]: fuse_norm_quant/fuse_act_quant custom fusions produce garbled token output on SM120 (GB10) for DeepSeek-V4-Flash-0731
- [Bug]: FLASHMLA_SPARSE_DSV4 backend crashes with "swa_metadata missing tile_sched entry for compress_ratio=1" on DeepSeek-V4-Flash-0731
- fix: Check all peers in _can_p2p when VLLM_SKIP_P2P_CHECK=1
- Docs
- Python not yet supported