Raullen Chai

Improve vllm-mlx performance through research, small experiments, and benchmarks on real models. The skill records results, keeps changes that improve measured performance, and carries successful work through implementation, testing, review, and pull request creation.

中文简介正在准备中。当前显示英语简介。

• Reducing the time before a model starts responding
• Improving token generation speed
• Exploring memory and cache optimizations