search

rllab-snu / Stage-Wise-CMORL

This is an official GitHub Repository for paper "Stage-Wise Reward Shaping for Acrobatic Robots: A Constrained Multi-Objective Reinforcement Learning Approach".
MIT License
64 stars 8 forks source link

after train and test student model ,what should we do next to test on actual robots, could you provide some references? #2

Open dbdxnuliba opened 3 weeks ago

dbdxnuliba commented 3 weeks ago

Hello, I have completed the training and testing of the student model about go1 backflip. What should I do next if I want to deploy the student model on a real go1 or do simulations in Mujoco? Could you please provide the deployment code related to go1 or provide some deployment ideas, thanks very much!

dbdxnuliba commented 3 weeks ago

@dobro12 hello, Could you provide some suggestions about simulation in mujoco or to real to export student actor policy

dbdxnuliba commented 2 weeks ago

@dobro12 could you please to have a look this quesiton,thanks very much

dobro12 commented 2 weeks ago

I am sorry for the late response.

Still, I am busy with lots of work whose deadlines are this week.

To briefly give my opinion in your case, I suggest copying and pasting the network codes, algos/common/*.* and algos/student/*.* into your local device. In our implementation, we don't save the network's structure but only save the weight/bias parameters.

Thank you.

Sincerely, Dohyeong Kim.

On Mon, Nov 11, 2024 at 4:48 PM dbdxnuliba @.***> wrote:

@dobro12 https://github.com/dobro12 could you please to have a look this quesiton,thanks very much

— Reply to this email directly, view it on GitHub https://github.com/rllab-snu/Stage-Wise-CMORL/issues/2#issuecomment-2467450087, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHUQNRAUPQGMHQLHR6LRX3T2ABOMRAVCNFSM6AAAAABQ2CKD3CVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRXGQ2TAMBYG4 . You are receiving this because you were mentioned.Message ID: @.***>

dbdxnuliba commented 2 weeks ago

Ok,thank you for your response. When you have time, could you please send me an example deployment code how to export weight/bias parameters and use them to the deployment project

dobro12 commented 1 week ago

I am sorry, but we do not have a plan to open our deployment code!🥲 Our suggestion is that you can refer to the "test" mode of the "main.py" and just replace the environment code with a Mujoco simulation or real-world robots.

dbdxnuliba commented 4 days ago

ok,I will have a try,thanks