Use RandomizeAction wrapper instead of Explorer in evaluation #328
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
muupan
changed the title
|
Resolved the conflict. |
| @@ -87,6 +88,9 @@ def make_env(process_idx, test): | |||
| episode_life=not test, | |||
| clip_rewards=not test) | |||
| env.seed(int(env_seed)) | |||
| if test: | |||
| # Randomize actions like epsilon-greedy in evaluation as well | |||
| env = chainerrl.wrappers.RandomizeAction(env, 0.05) | |||
There was a problem hiding this comment.
Is there a reason that you use 0.05 instead of args.eval_epsilon as in the other algorithms? I'm aware that the DQN paper uses 0.05, but then why not use a raw value for the other domains?
There was a problem hiding this comment.
I did so just because the previous examples/ale/train_nsq_ale.py did so. I agree that it's better to allow configuring it, but since this is not relevant to this PR, I kept it unchanged.
prabhatnagarajan
approved these changes
Oct 10, 2018
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We currently use
Explorerto inject randomness to action selection in evaluation, which is common in Atari benchmarks. This use is actually irrelevant to exploration in training, so is a misnomer in my opinion.I think we better use an env wrapper for this purpose, so that the training and evaluation code can be simpler. This is also important in #326 , where I need to implement another set of code for training and evaluation.