How can I adapt my own robotic arm and provide a rough adaptation process

[xiaoqigegeyou]

How can I adapt my own robotic arm? Can you provide a rough adaptation process? Additionally, there is a configuration issue with the Playground demo script, which does not automatically pull models from the huggingface and set single graphics card configurations.

[xiaoqigegeyou]

This should be set to 0 by default export HF_HUB_OFFLINE=0

[Zeng-Jia]

Based on the current code version (83b4827), you can add your robot_type by editing https://github.com/InternRobotics/InternVLA-A1/blob/master/src/lerobot/transforms/constants.py.

We’re also considering improving this part of the code to make it more user-friendly for adapting to new embodiments, and we’ll provide a tutorial as well.

[xiaoqigegeyou]

Thank you for the reminder. We have completed the fine-tuning phase in the demo, but found that the code for fine-tuning and evaluation does not match. They are somewhat independent. What should we do after receiving the chekpoints.

[xiaoqigegeyou]

The current inference only supports the cloud and is inconsistent with the local fine-tuning directory. Can you provide a script for local inference

[bbbbbbbbly]

how to use the embodiments or data in robotwin2 to finetune intervla-a1