Dysl-vla：Official code for "DySL-VLA: Efficient Vision-Language-Action Model Inference via Dynamic-Static Layer-Skipping for Robot Manipulation". A framework that dynamically skips VLA layers based on action importance, achieving 3.75× speedup and 85.7× training reduction on robotic manipulation tasks.

Research prototype: category-aware layer skipping for small LLMs on Apple Silicon. Honest write-up of what worked (1.9x batch throughput on Llama 3.2 3B) and what didn't (MMLU collapse). Outperformed by LayerSkip/MoD/vLLM.