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
- [Rust Frontend] Skip full OpenAI-server CLI schema for the managed headless engine
- [CPU] Enable GPTQ and AWQ quantization for s390x
- [Bugfix][Frontend][Kimi K3] Let /v1/responses disable spaces_between_special_tokens
- [Bugfix][Spec Decode] Bound draft max_model_len by the target before logging
- [Feature]: `vllm.utils.import_util._has_module` should not import modules
- [Bug]: cache_config_info reports block_size=4 despite --block-size 256, and num_gpu_blocks × block_size ≠ kv_cache_size_tokens (DeepSeek-V4, fp8_ds_mla)
- [Bug]: Step3p5ReasoningParser strips the wrong newline once the discarded prefix is preserved (follow-up to #50918)
- [Bugfix][MoE] Gate deepep_v2 prepare/finalize on cudagraph_mode, not enforce_eager
- [Model] Voxtral Realtime: add support for `CUDAGraphMode.FULL_DECODE_ONLY`
- [Rust Frontend] Fix Qwen reasoning parser auto-detection
- Docs
- Python not yet supported