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 supported59 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bugfix] Fix FA3 crash with kv_cache_dtype_skip_layers under an FP8 KV cache
- [ROCm][Bugfix] DeepSeek-V4: B-preshuffle the FP8 linears in their kernel so weight reloads redo it
- fix(sampler): fall back to the native sampler when FlashInfer cannot JIT-build
- [Perf][Mamba] Launch checkpoint export only for rows with a checkpoint
- [Kernel][Perf] Add Blackwell BF16 tuned configs for Qwen2.5-1.5B to the batch-invariant matmul table
- [Model] Move MiniMax-M3 to `AutoWeightsLoader`
- [HiSparse][DSv4.1] Host-back the C1/C2 compressed caches on SM90
- [Bug]: Qwen3.5-122B-A10B-FP8 GSM8K accuracy decreases at commit 0fae39a
- [Bug]: v0.33.0: DeepSeek-V4.1 + DSpark speculative decoding OOMs at startup with `--gpu-memory-utilization=0.97` (worked in v0.32.0) — speculator CUDA-graph-capture allocation (~4 GiB in `sparse_attn_indexer`) not accounted for by memory profiling
- [RFC]: Deprecate and remove the torchao reload decorator in favor of reload_weights
- Docs
- Python not yet supported