search

Daffan / ros_jackal

ROS-Jackal environment for RL
MIT License
28 stars 6 forks source link

Missing the td3/config.yaml. #4

Open happy-yasuo404 opened 1 year ago

happy-yasuo404 commented 1 year ago

Hello, Dr. Xu, when we tried to train TD3 using the applr_working branch of your code, we found that the training configurations were missing. There is no td3/config.yaml file under the td3 file, so I would be grateful if you could provide it generously.

Daffan commented 1 year ago

Hi, sorry I think it is a typo. Can you run python3 td3/train.py --config_path params_laser_local.yaml? This will train APPLR in one world locally (not on cluster), and it's very slow.

happy-yasuo404 commented 1 year ago

Dr. Xu, we have tried following your suggestion, as you said, this train APPLR in one world locally (not on cluster), and it's indeed very slow. How can we collect trajectories in a cluster (Condor Cluster for large relatively inaccessible to most users), or how do we collect data locally using multiple actors? We have read your algorithm and paper carefully and want to train the algorithm on our machine. Can you give us some help, thank you very much.