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
- [Bugfix] Support strided weights in sharded state saving
- [ROCm][Spec Decode] Rank unified attention above ROCM_ATTN for spec decode and hybrid models
- [Perf][Spec Decode] Defer GDN spec decode state stores to the next step
- [Kernel] Add FlashInfer GDN and KDA prefill backend selection
- [Spec Decode] Add draft_context_window to bound DFlash/DSpark draft attention
- [ROCm][Perf] Batch short TurboQuant verify blocks and enable split-KV
- [ROCm][AMD] Generalize MiniMax-M3 mono decode architecture for TP=2, TP=4, and TP=8 on gfx950
- [Build] Tag wheels cp311-abi3 and look up release wheel URLs in install docs
- [Bug]: LoRA MoE backend selection ignores --moe-backend and the TRT-LLM LoRA path asserts with --fully-sharded-loras
- [Frontend] Add BROWSER_BACKEND to select the gpt-oss demo browser backend
- Docs
- Python not yet supported