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 supported62 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [Performance]: Avoid runtime routed_scaling_factor multiply between all-reduce and GEMM (MoE)
- [Bug]: Fast Start WeightCacheKey hashes only safetensors headers, so an ipc_cache engine can silently map another checkpoint's weights
- GitHub API version 2022-11-28 is retired in March 2028 (run_ci_command.py)
- [Startup] Add --flashinfer-autotune-cache-only to load the FlashInfer autotune cache and skip startup autotuning
- [Feature][Spec Decode] ngram_hint: draft tool calls from the chat template's rendered call syntax
- [Feature]: Integrate Cake kernels via FlashInfer: model-by-model tracker
- [Doc] Document NVLS non-determinism on Hopper NVSwitch nodes
- [RFC]: Handle empty responses in synthetic acceptance benchmarks
- [Performance]: Nemotron-3.5-Lightning NVFP4 decode ~16% slower on DGX Spark (GB10/SM121) since v0.29.0
- [RFC]: Octave KV, a native 3-bit KV cache for AMD GPUs
- Docs
- Python not yet supported