Master Thesis project on Deep Reinforcement Learning.
- tensorflow 2.3.1
- tensorforce 0.6.2
- numpy 1.19.4
- matplotlib
- Step 1: Run tensorflow_train.py in order to train the agent.
- Step 2: After training, go in tensorforce_train.py line 26 and change the weights file name with the one you got from Step 1.
- Step 3: Check parameters in tensorforce_main.py (lines 17-23). Grid is set to be 2x2 by default.
- Step 4: Run tensorforce_train.py. The agent should start to train indefinitely on a single random image from CIFAR10.
Thanks @SestoAle for providing the tracker_dense.py script.
- PyTorch net implementation is currently not used, but it's working.