TRPO example that reproduces the "Deep Reinforcement Learning that Matters" paper#449
Merged
muupan merged 24 commits intochainer:masterfrom Sep 6, 2019
Merged
TRPO example that reproduces the "Deep Reinforcement Learning that Matters" paper#449muupan merged 24 commits intochainer:masterfrom
muupan merged 24 commits intochainer:masterfrom
Conversation
Member
Author
|
/test |
Collaborator
|
Successfully created a job for commit b0b017a: |
muupan
changed the title
toslunar
reviewed
Sep 6, 2019
|
|
||
| # Normalize observations based on their empirical mean and variance | ||
| obs_normalizer = chainerrl.links.EmpiricalNormalization( | ||
| obs_space.low.size, clip_threshold=5) |
Member
There was a problem hiding this comment.
I didn't find clip_threshold=5 in the paper, but 5 seems sufficiently large.
Member
Author
There was a problem hiding this comment.
They used a fork of openai/baselines. TRPO uses MlpPolicy, which internally clip normalized observations by 5.
https://github.com/Breakend/baselines/blob/master/baselines/trpo_mpi/run_mujoco.py#L31
https://github.com/Breakend/baselines/blob/master/baselines/ppo1/mlp_policy.py#L25
Co-Authored-By: Toshiki Kataoka <tos.lunar@gmail.com>
It should not affect the behavior as the default value of the argument is also 0.01.
Member
Author
|
/test |
Collaborator
|
Successfully created a job for commit 8762890: |
Member
Author
|
/test |
Collaborator
|
Successfully created a job for commit dfdcf5d: |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Merge
#446#448first.This PR adds a script to reproduce the TRPO results in "Deep Reinforcement Learning that Matters" (http://arxiv.org/abs/1709.06560).
As you can see from the tables in README, our results are competitive with theirs. https://github.com/muupan/chainerrl/tree/trpo-mujoco-matters/examples/mujoco/reproduction/trpo