Question about visualize the change of reward when training

matthieu637 / ddrl

Deep Developmental Reinforcement Learning

MIT License

29 stars 3 forks source link

Question about visualize the change of reward when training #7

Open huangjiancong1 opened 5 years ago

huangjiancong1 commented 5 years ago

Why we use test() to see the reward and testing_sample_step?

Can I use the train() to see how the reward change when training? It seems that the last perf is the reward.

Because we want to compare with the openai,

matthieu637 commented 5 years ago

The test() function disable exploration for testing the performance of deterministic policies. 0.0.monitor.csv contains training performance (with exploration) 0.1.monitor.csv contains testing performance (without exploration)