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 supported51 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- fix(harmony-renderer): cap system identity instructions at 4096 chars
- [Bug]: triton_w4a16_gemm assert qzeros.shape == (K // group_size, N // 8) fails on GPTQ models
- [BugFix][Frontend] Add content type validation for Harmony path (GPT-OSS models)
- [RoCm][Build/CI][The Rock] Fix test_gemma4router.py for The Rock due to unstable torch.topk
- [Build] Fix CUDA arch coverage checks and scoped kernel feature flags
- [Feature]: Support combining multiple speculative decoding methods (e.g. MTP + ngram)
- [Performance]: Audit RoPE + KV-cache fusion coverage for GQA, MLA, and SparseMLA on CUDA
- [Bug]: Why does kv-cache-dtype=fp8 OOM more easily than bf16/fp16 on long-context GLM-5.1-AWQ runs?
- [Bug]: AssertionError: Overwriting existing tensor attribute: weight_loader when serving FP8 model on 2x RTX 5090 (Blackwell)
- [CPU][Bug] Dynamic Speculative Decoding crashes on CPU backend
- Docs
- Python not yet supported