vllm

3 papers

• Controlling Thinking Speed in Reasoning Models
• Fast Inference for Augmented Large Language Models
• TransMLA: Migrating GQA Models to MLA with Full DeepSeek Compatibility and Speedup