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
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bug]: W4A8 (int4×FP8) MoE asserts hidden/intermediate %256 for every backend, so HUMMING cannot load shapes CUTLASS rejects (e.g. intermediate 640)
- [Bugfix] Let W4A8 MoE fall back to Humming for non-256-aligned shapes
- [Bugfix][Attention] Re-plan DSA indexer decode schedule for DBO's reduced DeepGEMM SM count
- [Bugfix][Attention] Keep DSA shared-indexer top-k state per DBO microbatch
- [Bug]: vLLM 0.31.0 Mooncake connector crashes after Decode timeout
- [Hybrid KV Cache] [3/N] Retain match-unit-aligned finalized Mamba decode checkpoints
- [Bug]: tool_choice="required" with the hermes structural tag drops parallel tool calls (separator forbids the template's newline)
- [Bugfix][Reasoning] Fix Granite streaming reasoning for multi-token deltas
- [Bugfix] add group count constraint and test cases for grouped topk M…
- [Performance]: Reduce load_weights synchronization overhead for DeepSeek V4 Flash weight updates on H200
- Docs
- Python not yet supported