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 supported23 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [Feature]: online bf16->fp8 (and maybe even fp4?) quantization in `load_weights(...)` (so not only from disk checkpoint, but importantly from VRAM or RAM) for weights reloading in GRPO rollout workloads, and speed gains for generating long reasoning rollouts and peak VRAM reduction
- [Bug]: 100% cpu usage on 3 cores on every node when using ray distributed pipeline parallel
- [Feature]: Any plans to add nvidia/parakeet-tdt-0.6b-v3 to vllm?
- [Usage]: "When running the QWEN3 MoE GGUF quantized model with VLLM, the inference output is all exclamation marks (!!!!!!!!!!!!!!!)"
- Change the default value of truncate_prompt_tokens in the embedding/rerank/pooling model to -1
- [Bug]: torch.AcceleratorError: CUDA error: an illegal memory access was encountered
- add log for request trace
- [Roadmap] vLLM Roadmap Q4 2025
- add log for request trace
- [Feature] Conditional prefix logging for stdout/stderr
- Docs
- Python not yet supported