How to finetune the model with single wrist image?

[jyC23333]

Hello!

I want to train the model with a dataset which is recorded by a single-arm robot. So I have only one wrist camera and one head camera.

1. If I just remove the other wrist image from the training, will the model still work after finetuning?
2. Was the model only trained on dual-arm datasets during the whole pretraining?
3. Do I need to pad the input images or output action_dims to adapt to the original go1 model?

[ModiShi]

1. Yes
2. Yes, GO-1 is only pretrained on AgiBot-World Beta
3. No