Code for my personal graduation paper

• Codes in TD3 are adapted from https://github.com/sfujim/TD3.git
• Codes in SAC are adapted from https://github.com/pranz24/pytorch-soft-actor-critic.git

Usage

Start training:

policy="adv"
# options: td3/adv
cd TD3/ && nohup sh run_experiments.sh ${policy} &
cd ../SAC/ && nohup sh run_experiments.sh ${policy} &

This will generate folders including logs, tensorboard, and train, containing training/testing logs, tensorboard log files, saved models after training, respectively.

Clean all temporary files

sh clean.sh

Run generalization tests:

nohup sh adv_test_all.sh &

This will generate log files in directory ./logs/eval_${name}_all.log, where name is either mass or noise, representing NT and HV tests respectively.