Open GXJll opened 11 months ago
Hi. Usually, it will take a day to finish training depending on how quick the convergence is. The robot seems to have converged to a poor policy.
I have found that the addition of local waypoints to the reward function can improve learning and prevent that problem. I'll be updating the repo soon, to fix that along with other small updates (rviz visualization of Collision Probability)
Hello. I've updated the readme instructions and repository to reflect the newer implementation I mentioned. Please let me know if it has some issues.
When running “roslaunch turtlebot3_rl_sim start_td3_training.launch” , there is an RunTimeError about size match. I think the code you have tested should not have size match problem. May I ask what other reasons do you think it might be?
Follow the steps of README, Can you run this project without GPU and CUDA?
When running “roslaunch turtlebot3_rl_sim start_td3_training.launch” , there is an RunTimeError about size match. I think the code you have tested should not have size match problem. May I ask what other reasons do you think it might be?
Can you paste the error logs?
Follow the steps of README, Can you run this project without GPU and CUDA?
Yes. The code can be run without GPU/Cuda.
Hi, I am facing the same issue, while executing the command ( roslaunch turtlebot3_rl_sim start_td3_training.launch ). In trackback, it shows a size mismatch for linear1.weight: copying a param with shape torch.Size([256, 363]) from checkpoint, the shape in the current model is a torch.Size([256, 398]). I think It means the number of input features (network_inputs) to the model is different at the time of training versus at the time of loading the model. Can you guide me regarding this issue?
I hope to get your reply.
I encountered the same problem---->RuntimeError: mat1 and mat2 shapes cannot be multiplied (1x398 and 363x256),Could you please tell me how to solve this problem?
Its size mismatch issues. I will help with that after an hour. If you have WhatsApp or WeChat let me know.
On Mon, May 20, 2024 at 19:28 Heritagecll @.***> wrote:
I encountered the same problem---->RuntimeError: mat1 and mat2 shapes cannot be multiplied (1x398 and 363x256),Could you please tell me how to solve this problem?
— Reply to this email directly, view it on GitHub https://github.com/ailabspace/drl-based-mapless-crowd-navigation-with-perceived-risk/issues/2#issuecomment-2120258239, or unsubscribe https://github.com/notifications/unsubscribe-auth/AVSDL5WL7RPFCQOO2X7CSQLZDHM5JAVCNFSM6AAAAAA5D5EBK2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMRQGI2TQMRTHE . You are receiving this because you commented.Message ID: <ailabspace/drl-based-mapless-crowd-navigation-with-perceived-risk/issues/2/2120258239 @github.com>
Thank you very much for your help,my WeChat is “cll812088140”,Could you please contact me? Thank you again!
I have also encountered a size mismatch issue. Can you tell me how to solve it
Hi @wolf212121j, can you paste your error here?
Its size mismatch issues. I will help with that after an hour. If you have WhatsApp or WeChat let me know. … On Mon, May 20, 2024 at 19:28 Heritagecll @.***> wrote: I encountered the same problem---->RuntimeError: mat1 and mat2 shapes cannot be multiplied (1x398 and 363x256),Could you please tell me how to solve this problem? — Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AVSDL5WL7RPFCQOO2X7CSQLZDHM5JAVCNFSM6AAAAAA5D5EBK2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMRQGI2TQMRTHE . You are receiving this because you commented.Message ID: <ailabspace/drl-based-mapless-crowd-navigation-with-perceived-risk/issues/2/2120258239 @github.com>
Oh,brother,I really need your help,my WeChat is "15656790801",can you add me ?
Hello, how long does this take to train? What is the reason why the robot is still spinning in the same place after I have trained for a day or so? I hope to get your reply.