DDPG fixes #75
Closed
DDPG fixes #75
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
Owner
kengz
commented
Mar 20, 2017
•
kengz
commented
- make DDPG work for cartpole first, for sanity check
- try dis https://github.com/songrotek/DDPG/blob/master/actor_network.py
- and try dis https://github.com/pemami4911/deep-rl/blob/master/ddpg/ddpg.py
- batch normalization
- make DDPG work for pendulum
- schedule DDPG for all the experiments
- check random seeds
lgraesser
reviewed
Mar 25, 2017
rl/policy/noise.py
Outdated
| @@ -44,7 +44,7 @@ def select_action(self, state): | |||
| Q_state = agent.actor.predict(state)[0] | |||
| assert Q_state.ndim == 1 | |||
| action = np.argmax(Q_state) | |||
| logger.info(str(Q_state)+' '+str(action)) | |||
| # logger.info(str(Q_state)+' '+str(action)) | |||
| return action | |||
|
|
|||
| def update(self, sys_vars): | |||
There was a problem hiding this comment.
AnnealedGaussian calls self.sample() but I don't think sample has been defined.
Also, how to explore with discrete actions. Add an epsilon greedy component to the policy?
|
I think it's really close to working but something is off with the gradient update. The DDPG implementation from permami works, and ddpg_tf implementation results in exactly the same actions up to the first gradient update. I also did some testing around the weights init - the algorithm is not that sensitive to them and will work for different initializations. |
|
working DDPG implemented in #118 |
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Outdated suggestions cannot be applied.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.