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 supported31 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bug]: Cross-node CUDA graph capture fails (illegal memory access at capture_end) — host-staged NCCL all-reduce captured by `breakable_cudagraph` on GB10 (no GPUDirect); `splitting_ops` not honored
- [Bug]: deepseek_v4 / deepseek_v32 tokenizers ignore add_generation_prompt and continue_final_message (silent empty/garbled output on assistant-terminated conversations)
- [Bug]: ModelOpt NVFP4 checkpoint causes fixed ~52 GiB CPU RAM allocation, OOM on 64GB system even for 0.8B model
- [ROCm][Perf][DSV4] Enable split sparse decode on gfx942
- [BugFix] weights processing peak memory reduction for nvfp4 MoE layers
- [Installation]: Fans spinning at 100% if no request is being processed on dual ARC B60 XPU setup
- [V1][KVConnector] Reclaim KV blocks leaked by aborted remote KV recv
- [Bugfix][Quant] Raise actionable error instead of bare assert for group-size/TP mismatch (#46230)
- [Bugfix] Fix type-sensitive tool argument coercion
- [Bugfix] DeepSeek-V4 tokenizer: honor add_generation_prompt and continue_final_message
- Docs
- Python not yet supported