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 supported3 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [Bug]: Logprob values are affected by sampling parameters and are incompatible with OpenAI API
- [Usage]: Regarding VLLM Structured Output Doubts
- [Bug]: Error with structured output inference after upgrade 0.6.2->0.6.3
- [Bug]: Too Many Tokens are Empty Strings and Empty Bytes, and `top_logprobs` Can't Identify End-of-Text (EOT) Tokens
- [Feature]: Alternating local-global attention layers
- [Misc]: [Question] vLLM's model loading & instance contract, 1 model per vLLM instance, or multiple models per vLLM instance
- [Bug]: Speculative decoding generate gibberish when receiving parallel requests with different seeds
- [Usage]: After upgrading vLLM from version 0.5 to 0.6, the FastChat vllm_worker fails to load model weights when used with the openai_server deployment method. This issue may be related to vLLM 0.6 requiring pydantic > 2.9. Is this compatibility issue caused by FastChat 0.2.36 not being updated recently, making it unable to adapt to vLLM 0.6?
- [Installation]: No module named 'vllm._version' from vllm.version import __version__ as VLLM_VERSION
- [Performance]: Maximizing the performance of batch inference of big models on vllm 0.6.3
- Docs
- Python not yet supported