Testing Issue

[henrytsui000]

Hi, I want to run this repo on my robot, but I face some problems:

After change the Robot-class, Could I use the model weights without re-train? I found that every-time I change Robot-class I need to use $pip install e . And need to retrain the model, otherwise it would response "ValueError: Value network is not well trained."

Does it have easy way to use own input? I have collect the real-time human data, but I don't know how to use these data except write a new test.py

[ChanganVR]

Sorry for the late reply.

I'm not sure if I understand your problem correctly. Are you asking if you can test a different policy with the same test.py? This should be totally doable and you shouldn't need to run pip install -e . for testing different policies unless you are doing something different.

If you are asking about running the policy with real-time data, yes, you would need to write a different test that accomodate your needs. But the model weights should be loaded just fine.