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 supported51 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Rust Frontend] Add Step3p5 tool parser
- [Bugfix][ROCm][Disagg] Honor pinned remote_dp_rank for WideEP child-pod notify
- [Bugfix][KVConnector] Drop unexpected send completions in MultiConnector
- [RFC]: Internal State Checkpoints for Mamba Align Mode
- [Bug][ROCm] Kimi-K3 (KimiLinear/GDN hybrid) + ngram spec-decode: GPU memory access fault on concurrent mixed spec/non-spec batches (single-stream works after fixing a strided `state_indices` slice)
- [Feature][ROCm] Per-layer-type KV-scale calibration for hybrid models — calibrate full-attention layers only, skip recurrent layers (unblocks fp8 KV on GDN/KimiLinear hybrids)
- Make Laguna MoE router scoring config-driven
- [Bugfix][Core] Drop stale block hashes when a session truncates tokens
- [Test] Cover DSpark SM120 sparse MLA topk widths
- [Bug]: benchmark_moe.py --tune --dtype int4_w4a16 crashes on NVIDIA (BLOCK_SIZE_K vs group_size)
- Docs
- Python not yet supported