Merge mw v2 examples #2287
Closed
Merge mw v2 examples #2287
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
added 7 commits
June 13, 2021 14:41
ML1 launchers and ML10 launchers Add rl2 with stddev clipping and neg log liklihood entroy and shared std and mean network. Use ryan's tanh clipping in the gaussian gru model/policy to maintain stability in the KL_before metric. Use a Linear baseline in MAML instead of the a mlp baseline, which had trouble fitting to observations between the different tasks All MT10 Launchers TE update to log named params Modify Pearl to add logging of train tasks add pearl train logging and also increase the train tasks to be large enough for ML10 in pearl_ml10 launcher Update Launchers Metaworld v1 mt10 experiments Modify launchers with final hparams and launching rate limiter RL2 tuned for old metaworld ML10 launchers for v1 environments ML10 old v1 launchers Final V2 ML10 MT10 launchers Launchers Used for final MT10 and ML10 Results Launchers round3 Updated door envs in metaworld for ML10 environments Final ML10 MT10 Launchers used for the Paper Initial Commit MT1 launchers Extend RL2 MTPPO MTTRPO experiments update metaworld to master ML1 rl2 round2 Increase rl2 ppo ml10 steps Rl2 Ml1 fixes decrease ml1-rl2-ppo eval epochs Reduce Logging epochs for rl2_ppo ml10 MT10 5e8 steps Downgrade torch TE PPO increase timesteps MT1 Experiments with fixed MT1 Adjust TE to log all tasks Switch back to epoch cycles Update MTPPO and MTTRPO params MTPPO MT50 MTTRPO MT50 launcher TE PPO MT50 MTSAC MT50 MAML and RL2 ML45 New hparams ml1 PEARL Debugging MAML ML45
mergify
bot
requested review from
a team,
gitanshu and
haydenshively
and removed request for
a team
- remove the torch linear feature baseline wrapper - add param docs to all new functions and launchers - delete unused docker and gcp launch files - mt1 examples do not work yet
|
The MT1 examples don't work at the moment, but everything else is looking good. @krzentner can you ptal? |
Hey @avnishn! Thanks for your efforts! What do you mean by this statement? Is there a bug in MT1 training examples? (eg : mtppo_metaworld_mt1_push.py) |
|
Anyone looking to reproduce the Meta-World results with garage should checkout the code in this PR. The following command should do the trick: |
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.
This is a PR in progess ... definitely doesn't pass pre commit, test cov, etc. There are a few files that have been added that shouldn't have been, it would be nice if @krzentner could take a look at the end of the week once I've made some more progress on cleaning up these files.