ikostrikov/pytorch-a2c-ppo-acktr-gail issues and pull requests

#100 - Bad performance on Humanoid-v2 using PPO

Issue - State: closed - Opened by lyp741 over 6 years ago - 2 comments

#100 - Bad performance on Humanoid-v2 using PPO

Issue - State: closed - Opened by lyp741 over 6 years ago - 2 comments

#99 - Naive question: why there is no envs.reset() in main.py?

Issue - State: closed - Opened by pengzhenghao over 6 years ago - 2 comments

#99 - Naive question: why there is no envs.reset() in main.py?

Issue - State: closed - Opened by pengzhenghao over 6 years ago - 2 comments

#98 - Fixes bug in recurrent policy. Issue #97

Pull Request - State: closed - Opened by erikwijmans over 6 years ago - 1 comment

#98 - Fixes bug in recurrent policy. Issue #97

Pull Request - State: closed - Opened by erikwijmans over 6 years ago - 1 comment

#97 - Bug in recurrent policy

Issue - State: closed - Opened by erikwijmans over 6 years ago - 1 comment

#97 - Bug in recurrent policy

Issue - State: closed - Opened by erikwijmans over 6 years ago - 1 comment

#96 - Termination gradient fix

Pull Request - State: closed - Opened by ronsailer over 6 years ago

#96 - Termination gradient fix

Pull Request - State: closed - Opened by ronsailer over 6 years ago

#95 - Question about PPO

Issue - State: closed - Opened by 0xsamgreen over 6 years ago - 3 comments

#95 - Question about PPO

Issue - State: closed - Opened by 0xsamgreen over 6 years ago - 3 comments

#94 - shared weights

Issue - State: closed - Opened by mkbera over 6 years ago - 2 comments

#94 - shared weights

Issue - State: closed - Opened by mkbera over 6 years ago - 2 comments

#93 - MLP recurrent

Pull Request - State: closed - Opened by timmeinhardt over 6 years ago - 9 comments

#93 - MLP recurrent

Pull Request - State: closed - Opened by timmeinhardt over 6 years ago - 9 comments

#92 - training parameter

Issue - State: closed - Opened by ghost over 6 years ago - 1 comment

#92 - training parameter

Issue - State: closed - Opened by ghost over 6 years ago - 1 comment

#91 - training

Issue - State: closed - Opened by ghost over 6 years ago - 1 comment

#91 - training

Issue - State: closed - Opened by ghost over 6 years ago - 1 comment

#90 - updating repo

Pull Request - State: closed - Opened by ronsailer over 6 years ago - 2 comments

#90 - updating repo

Pull Request - State: closed - Opened by ronsailer over 6 years ago - 2 comments

#89 - GRU doesn't work for A2C

Issue - State: closed - Opened by ShaniGam over 6 years ago - 7 comments

#89 - GRU doesn't work for A2C

Issue - State: closed - Opened by ShaniGam over 6 years ago - 7 comments

#88 - CartPole-v0 reward above 200.

Issue - State: closed - Opened by codeislife99 over 6 years ago - 4 comments

#88 - CartPole-v0 reward above 200.

Issue - State: closed - Opened by codeislife99 over 6 years ago - 4 comments

#87 - VecNormalize gamma parameter

Issue - State: closed - Opened by timmeinhardt over 6 years ago - 2 comments

#87 - VecNormalize gamma parameter

Issue - State: closed - Opened by timmeinhardt over 6 years ago - 2 comments

#86 - Add assert statements for PPO data generator batch sizes

Pull Request - State: closed - Opened by timmeinhardt over 6 years ago - 1 comment

#86 - Add assert statements for PPO data generator batch sizes

Pull Request - State: closed - Opened by timmeinhardt over 6 years ago - 1 comment

#85 - How to make K-FAC work with BatchNorm layer?

Issue - State: closed - Opened by knn1989 over 6 years ago - 2 comments

#85 - How to make K-FAC work with BatchNorm layer?

Issue - State: closed - Opened by knn1989 over 6 years ago - 2 comments

#84 - Added gamma parameter to VecNormalize init

Pull Request - State: closed - Opened by KMarino over 6 years ago

#84 - Added gamma parameter to VecNormalize init

Pull Request - State: closed - Opened by KMarino over 6 years ago

#83 - Use builtin loss function

Pull Request - State: closed - Opened by alok over 6 years ago - 1 comment

#83 - Use builtin loss function

Pull Request - State: closed - Opened by alok over 6 years ago - 1 comment

#82 - Question about PPO recurrent policy

Issue - State: closed - Opened by gliese581gg over 6 years ago - 1 comment

#82 - Question about PPO recurrent policy

Issue - State: closed - Opened by gliese581gg over 6 years ago - 1 comment

#81 - Apply SplitBiases in KFAC only once to a model

Pull Request - State: closed - Opened by timmeinhardt over 6 years ago

#81 - Apply SplitBiases in KFAC only once to a model

Pull Request - State: closed - Opened by timmeinhardt over 6 years ago

#80 - could you upload Walker2d-v2 and Humanoid-v2 PPO pretrained model?

Issue - State: open - Opened by jiameij over 6 years ago - 5 comments

#80 - could you upload Walker2d-v2 and Humanoid-v2 PPO pretrained model?

Issue - State: open - Opened by jiameij over 6 years ago - 5 comments

#79 - ppo recurrent check fixed.

Pull Request - State: closed - Opened by dineshj1 over 6 years ago - 1 comment

#79 - ppo recurrent check fixed.

Pull Request - State: closed - Opened by dineshj1 over 6 years ago - 1 comment

#78 - Meaning on value_preds and returns.

Issue - State: closed - Opened by cmuspencerlo over 6 years ago - 2 comments

#78 - Meaning on value_preds and returns.

Issue - State: closed - Opened by cmuspencerlo over 6 years ago - 2 comments

#77 - MLP policy recurrent version

Pull Request - State: closed - Opened by dineshj1 over 6 years ago

#77 - MLP policy recurrent version

Pull Request - State: closed - Opened by dineshj1 over 6 years ago

#76 - Make entire project a python package and installable with pip

Pull Request - State: closed - Opened by timmeinhardt over 6 years ago - 6 comments

#75 - Improve readability of policy act method

Pull Request - State: closed - Opened by timmeinhardt over 6 years ago - 1 comment

#75 - Improve readability of policy act method

Pull Request - State: closed - Opened by timmeinhardt over 6 years ago - 1 comment

#74 - Deterministic Policy

Issue - State: closed - Opened by timmeinhardt over 6 years ago - 4 comments

#74 - Deterministic Policy

Issue - State: closed - Opened by timmeinhardt over 6 years ago - 4 comments

#73 - Multithreading with PPO

Issue - State: closed - Opened by tshrjn over 6 years ago - 1 comment

#73 - Multithreading with PPO

Issue - State: closed - Opened by tshrjn over 6 years ago - 1 comment

#72 - ValueError if batch size is smaller than number of mini batches

Issue - State: closed - Opened by timmeinhardt over 6 years ago - 1 comment

#72 - ValueError if batch size is smaller than number of mini batches

Issue - State: closed - Opened by timmeinhardt over 6 years ago - 1 comment

#71 - Match code to pretrained models

Issue - State: open - Opened by chrirupp over 6 years ago - 3 comments
Labels: bug

#71 - Match code to pretrained models

Issue - State: open - Opened by chrirupp over 6 years ago - 3 comments
Labels: bug

#70 - removing wrong import in ppo

Pull Request - State: closed - Opened by prolearner over 6 years ago - 1 comment

#69 - How to use pixel training instead of low-dimensional state?

Issue - State: closed - Opened by knn1989 over 6 years ago - 6 comments

#68 - Suggestion: Mujoco - add timestep to the observation

Issue - State: closed - Opened by tesslerc over 6 years ago - 4 comments

#67 - Missing `args.value_loss_coef`?

Issue - State: closed - Opened by lcswillems over 6 years ago - 2 comments

#66 - The reward become negative soon and can't train the model well

Issue - State: closed - Opened by mxmxlwlw over 6 years ago - 1 comment

#65 - LSTM policy hyper parameters

Issue - State: closed - Opened by playerkk over 6 years ago - 2 comments

#64 - Could you please share the hyper-parameters for KUKA use PPO?

Issue - State: closed - Opened by LizstCat over 6 years ago - 3 comments

#63 - PPO episode ends before num_steps

Issue - State: closed - Opened by svd3 over 6 years ago - 2 comments

#62 - make plots scale to fit num_frames

Pull Request - State: closed - Opened by willwhitney over 6 years ago - 1 comment

#61 - Add support for DeepMind Control Suite

Pull Request - State: closed - Opened by willwhitney over 6 years ago - 3 comments

#60 - Could you share the hyper-parameters of a2c with different discrete action space

Issue - State: closed - Opened by BoyuanYan over 6 years ago - 1 comment

#59 - Custom Env log problems

Issue - State: closed - Opened by heitorrapela over 6 years ago - 2 comments

#58 - Missing Documentation Plotting Script

Issue - State: closed - Opened by araffin over 6 years ago - 2 comments

#57 - K-FAC

Issue - State: closed - Opened by lukashermann over 6 years ago - 1 comment

#56 - What's the shape of the input param actions in logprobs_and_entropy in distributions.py?

Issue - State: closed - Opened by BoyuanYan over 6 years ago - 4 comments

#55 - large negative rewards for Pong

Issue - State: closed - Opened by ammirato over 6 years ago - 6 comments

#54 - potential bugs in kfac.py

Issue - State: open - Opened by gd-zhang over 6 years ago - 1 comment
Labels: help wanted

#53 - A2C performace on Seaquest

Issue - State: closed - Opened by gautamb85 over 6 years ago - 4 comments

#52 - Two questions regarding recurrent policies

Issue - State: closed - Opened by maximecb almost 7 years ago - 3 comments

#51 - Fix minor bug wrt recurrent policy

Pull Request - State: closed - Opened by maximecb almost 7 years ago

#50 - Upload pre-trained models for A2C

Issue - State: closed - Opened by gautamb85 almost 7 years ago

#49 - GAE implementation for PPO

Issue - State: closed - Opened by wjaskowski almost 7 years ago - 1 comment

#48 - Why are the value_loss and action_loss summed?

Issue - State: closed - Opened by lintangsutawika almost 7 years ago - 2 comments

#47 - Is there a way to track the episode length?

Issue - State: closed - Opened by bearpaw almost 7 years ago - 1 comment

#46 - add explict dimension choice for softmax and log-softmax function

Pull Request - State: closed - Opened by bearpaw almost 7 years ago - 1 comment

#45 - Add argument to select the port to run the visdom server on

Pull Request - State: closed - Opened by bearpaw almost 7 years ago - 1 comment

#44 - Support for MultiDiscrete action spaces

Issue - State: closed - Opened by timmeinhardt almost 7 years ago - 5 comments
Labels: help wanted

#43 - Catch small bin_size parameters

Issue - State: closed - Opened by timmeinhardt almost 7 years ago - 1 comment

#42 - Why do we need to evaluate the actor_critic model twice?

Issue - State: closed - Opened by maximilianigl almost 7 years ago - 2 comments

#41 - Update to run roboschool,

Pull Request - State: closed - Opened by mabirck almost 7 years ago - 1 comment

#40 - Added ScaleObservations environment wrapper

Pull Request - State: closed - Opened by maximecb almost 7 years ago - 2 comments

#39 - remove moving stats to cpu.

Pull Request - State: closed - Opened by hengyuan-hu almost 7 years ago - 3 comments

#38 - Hyperparameter for mujoco environment

Issue - State: closed - Opened by a7b23 almost 7 years ago - 1 comment

#37 - log showing making new env twice

Issue - State: closed - Opened by CarloLucibello almost 7 years ago - 1 comment

#36 - broken master (input shape)

Issue - State: closed - Opened by CarloLucibello almost 7 years ago - 1 comment

#35 - typo in arg

Issue - State: closed - Opened by CarloLucibello almost 7 years ago - 1 comment

#34 - Added linear schedule for learning rate

Pull Request - State: closed - Opened by linusericsson almost 7 years ago - 1 comment

#33 - Using MLP model for small inputs

Pull Request - State: closed - Opened by maximecb almost 7 years ago - 4 comments

#32 - Make WrapPyTorch apply to non-atari environments

Pull Request - State: closed - Opened by maximecb almost 7 years ago - 1 comment

#31 - Linear schedule for learning rate

Issue - State: closed - Opened by linusericsson almost 7 years ago - 4 comments

#30 - default load_dir in enjoy.py is incorrect

Issue - State: closed - Opened by maximecb almost 7 years ago - 4 comments

GitHub / ikostrikov/pytorch-a2c-ppo-acktr-gail issues and pull requests