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
- [Bug]: bare assert in BasevLLMParameter._assert_and_load gives no diagnostics on weight shape mismatch
- [Bug]: LinearBase.load_weights substitutes the module for a missing parameter, surfacing as a confusing AttributeError
- [Kernel] Support head_dim=512 in fused QK-norm+RoPE (enables Gemma4)
- fix: surface clear error for missing parameters in LinearBase weight loader
- fix: populate draft model quant mapping and guard DFlash fused KV
- [Bug]: [Bug]: bare `@torch.compile` in `kimi_k25_vit` compiles outside the compilation lifecycle and pins `TRITON_CACHE_DIR` process-wide
- attn res + sigmoid_mul + conv fusions
- [ROCm][aiter] officially supporting aiter Triton kernels+dsv4 on RDNA3
- [Bugfix] Prevent stale partial prefix-cache hash resurrection after full-block promotion
- [Bugfix][ROCm] Fix num_tokens inflation for 2-D (M-RoPE) positions from the MTP drafter
- Docs
- Python not yet supported