InstructVLA
Collection
Paper, Data and Checkpoints for ``InstructVLA: Vision-Language-Action Instruction Tuning from Understanding to Manipulation''
•
14 items
•
Updated
•
1
Eagle2-2B: Uses the same pretrained weights as https://huggingface.co/nvidia/Eagle2-2B, with a modified modeling_eagle_chat.py
for bug fixes.
empty_language_adapter: An empty LoRA adapter for Eagle2-2B, used in Stage-2 training for InstructVLA.
Please move these two folders to path_to_project_root/ckpt
.
Base model
nvidia/Eagle2-2B