Features/openai hacks #35
Open
Conversation
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
…ork on pusher and pick and place (probably slider too but didnt test) 1. Online observation normalization and clipping 2. Adding additional loss term penalizing large preactivations in the policy network 3. Clipping Q-values in the Q network to stay within semantic return bounds
vitchyr
reviewed
Feb 22, 2019
| import numpy as np | ||
|
|
||
|
|
||
| class GaussianAndEpislonStrategy(RawExplorationStrategy, Serializable): |
rlkit/torch/networks.py
Outdated
| self.composite_normalizer = composite_normalizer | ||
|
|
||
| def forward(self, obs, **kwargs): | ||
| if self.composite_normalizer: |
There was a problem hiding this comment.
This check seems a bit redundant given the assert statement in __init__.
rlkit/torch/networks.py
Outdated
| def __init__( | ||
| self, | ||
| *args, | ||
| composite_normalizer: CompositeNormalizer = None, |
There was a problem hiding this comment.
Seems like we can just make this a required argument rather than kwarg.
rlkit/torch/td3/td3.py
Outdated
| if self.clip_q: | ||
| target_q_values = torch.clamp( | ||
| target_q_values, | ||
| -1/(1-self.discount), |
There was a problem hiding this comment.
Can you make this a parameter rather than hard-coding it? It could be something like:
if max_q_value is None:
max_q_value = -1/(1-self.discount) # for HER sparse rewards.
in __init__.
…to features/openai-hacks
|
I changed it to clip the networks instead, this should be easier. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.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.
Sorry for the delay...
These are the changes for successful pushing/pick and place in state space + typo in gaussian epsilon
I tried to cleanup the commit history with git rebase -i ec231e7 but it's behaving strangely; after I run that command, it pulls a much longer history of the commits, including some from the main branch.