Evaluate with tensorboard

There is automatic evaluation and logging in the train script. It is controlled by parameters –eval_freq and –eval_episodes. The log files are stored in the folder with trained model and you can easily visualize learning progress after the training and compare training parameters. There are also gifs for each eval period stored to compare robot performance during training. We also implemented evaluation in tensorboard:

tensorboard --logdir ./trained_models/yourmodel