search

Improbable-AI / walk-these-ways

Sim-to-real RL training and deployment tools for the Unitree Go1 robot.
https://gmargo11.github.io/walk-these-ways/
Other
488 stars 129 forks source link

How to run a newly trained network on the Go1? #32

Closed Robokan closed 1 year ago

Robokan commented 1 year ago

I have trained a new network and want to deploy it onto my Go1. It looks to be using rsync to move all the runs to the go1. How do I get it to choose my run rather than pretrain-v0? (which I am assuming it automatically uses).

Robokan commented 1 year ago

Found it. Line 73 in go1_gym_deploy/scripts/deploy_policy.py

Qianzhong-Chen commented 1 year ago

Hi, could you please be so kind to tell me how to make the self-trained policy (a .pt file) be available to be deployed with code of this repo? If my guess is correct, deploy_policy.py mainly uses the two .jit files (adaptation_module_latest.jit and body_latest.jit).

Robokan commented 1 year ago

I haven't done it yet but as far as I can tell all you do is 1) train a new network. Then the result will be in the "runs" directory. Then

2) edit: deploy_policy.py to use your run rather than the pretrained one.

3) follow the instructions to deploy to the go1. One of the scrips contains rsync I believe it will copy all the runs over to the go1 (including yours)