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 supported59 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Quantization] feat: add support for lossless NF12 online quantization
- [Bugfix][EPLB] Report async EPLB worker failures to fault tolerance instead of stalling
- [Bugfix][ROCm] benchmark_moe: keep HIP_VISIBLE_DEVICES for Ray
- [Bugfix][Tool Parser] Olmo3: stream plain text that starts with "<"
- [Bug][KV Connector] Permanent stall: parked async KV loads exhaust the GPU pool and none can allocate its tail (OffloadingConnector, spec decode on, 0.31-0.33)
- [Bugfix][Mooncake] Bound store prefix lookups with a circuit breaker
- [doc] three core functions document a return value they cannot produce (moe_unpermute, CacheOnlyModel.forward, InputBatch.condense)
- [Bug]: ModelOpt FP8 KV-cache scales are not loaded into MLA attention (DeepSeek-V2/V3, GLM-4.7-Flash, A.X-K1, EAGLE-3 DeepSeek drafter) since v0.25.0
- [Bug]: GPT-J silently ignores FP8 KV-cache scales from compressed-tensors (LLM Compressor) checkpoints
- [Bug]: Kimi-Linear ignores checkpoint FP8 KV-cache scales (compressed-tensors and ModelOpt): its MLA layer gets no KV-cache quant method
- Docs
- Python not yet supported