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
Add a CodeTriage badge to vllm
Help out
- Issues
- [Feature]: Unify activation-scheme handling across linear oracles (generalize the activation override + add a uniform selection gate)
- [Bugfix][Quantization] Select batch-invariant MXFP8/MXFP4 linear kernels under VLLM_BATCH_INVARIANT
- [RFC][Prototype] Snapshot post-processed model weights to skip checkpoint loading on restart
- [Bugfix][Model] Build the tied embedding on the last PP rank for ERNIE 4.5 MoE and Falcon-H1
- [Bugfix][Spec Decode] Fix standalone draft models with pipeline parallelism in Model Runner V2
- [ROCm][Muse Spark] DSpark draft layers as one persistent FlyDSL launch (--monokernel)
- [Bugfix] Make fused MoE expert weights contiguous before GPU copy in Llama-4
- [Doc] Add a new-model checklist to the parser agent guide
- [Bugfix][ROCm] Support AITER's preload() in FlyDSL INT4 QuickReduce init
- [Bugfix][MRV2] Build multimodal prefix-LM attention ranges on every PP stage
- Docs
- Python not yet supported