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 supported4 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [Bugfix][Kernel] Give unique name to BlockSparseFlashAttention
- [Bug]: torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 224.00 MiB. GPU 0 has a total capacity of 11.53 GiB of which 187.75 MiB is free. Including non-PyTorch memory, this process has 11.34 GiB memory in use.
- [Usage]: Failed to serve local model in distributed inference
- [Feature]: Allow setting a max_tokens (max_completion_tokens in OpenAI API) for all requests.
- [Bug]: Issue with Multi-Node Multi-GPU Deployment of DeepSeek-v2.5 on NVIDIA 4090
- [Feature]: Phi-4 tool support
- [Bug]: CPU Offloading errors (Worker.__init__() got an unexpected keyword argument 'kv_cache_dtype')
- [Doc]: Can the '/metrics' interface be disabled through configuration
- [Installation]: `numpy < 2.0.0` is annoying. Pls support `numpy == 2`
- [Bug]: The api server /health endpoint is unable to detect when the Worker VllmWorkerProcess has died
- Docs
- Python not yet supported