Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/araffin/srl-zoo

State Representation Learning (SRL) zoo with PyTorch - Part of S-RL Toolbox
https://github.com/araffin/srl-zoo

autoencoder deep-learning forward-model inverse-model neural-network pytorch reinforcement-learning representation-learning srl state-representation-learning vae

Last synced: about 1 month ago
JSON representation

State Representation Learning (SRL) zoo with PyTorch - Part of S-RL Toolbox

Awesome Lists containing this project

README 

        
# State Representation Learning Zoo with PyTorch (part of S-RL Toolbox)



A collection of State Representation Learning (SRL) methods for Reinforcement Learning, written using PyTorch.






SRL Zoo Documentation: https://srl-zoo.readthedocs.io/



S-RL Toolbox Documentation: https://s-rl-toolbox.readthedocs.io/



S-RL Toolbox Repository: https://github.com/araffin/robotics-rl-srl



Available methods:



- Autoencoder (reconstruction loss)

- Denoising Autoencoder (DAE)

- Forward Dynamics model

- Inverse Dynamics model

- Reward prediction loss

- Variational Autoencoder (VAE) and beta-VAE

- SRL with Robotic Priors + extensions (stereovision, additional priors)

- Supervised Learning

- Principal Component Analysis (PCA)

- Triplet Network (for stereovision only)

- Combination and stacking of methods

- Random Features

- **[experimental]** Reward Prior, Episode-prior, Perceptual Similarity loss (DARLA), Mutual Information loss



Related papers:

- "Decoupling feature extraction from policy learning: assessing benefits of state representation learning in goal based robotics" (Raffin et al. 2018) [https://arxiv.org/abs/1901.08651](https://arxiv.org/abs/1901.08651)

- "S-RL Toolbox: Environments, Datasets and Evaluation Metrics for State Representation Learning" (Raffin et al., 2018) [https://arxiv.org/abs/1809.09369](https://arxiv.org/abs/1809.09369)

- "State Representation Learning for Control: An Overview" (Lesort et al., 2018), link: [https://arxiv.org/pdf/1802.04181.pdf](https://arxiv.org/pdf/1802.04181.pdf)






## Documentation



Documentation is available online: [https://srl-zoo.readthedocs.io/](https://srl-zoo.readthedocs.io/)



## Installation



Please read the [documentation](https://s-rl-toolbox.readthedocs.io/) for more details, we provide anaconda env files and docker images.



## Learning a State Representation



To learn a state representation, you need to enforce constrains on the representation using one or more losses. For example, to train an autoencoder, you need to use a reconstruction loss.

Most losses are not exclusive, that means you can combine them.



All losses are defined in `losses/losses.py`. The available losses are:



- autoencoder: reconstruction loss, using current and next observation

- denoising autoencoder (dae): same as for the auto-encoder, except that the model reconstruct inputs from

                             noisy observations containing a random zero-pixel mask

- vae: (beta)-VAE loss (reconstruction + kullback leiber divergence loss)

- inverse: predict the action given current and next state

- forward: predict the next state given current state and taken action

- reward: predict the reward (positive or not) given current and next state

- priors: robotic priors losses (see "Learning State Representations with Robotic Priors")

- triplet: triplet loss for multi-cam setting (see *Multiple Cameras* section in the doc)



**[Experimental]**

- reward-prior: Maximises the correlation between states and rewards (does not make sense for sparse reward)

- episode-prior: Learn an episode-agnostic state space, thanks to a discriminator distinguishing states from same/different episodes

- perceptual similarity loss (for VAE):  Instead of the reconstruction loss in the beta-VAE loss, it

uses the distance between the reconstructed input and real input in the embedding of a pre-trained DAE.

 - mutual information loss: Maximises the mutual information between states and rewards



All possible arguments can be display using `python train.py --help`. You can limit the training set size (`--training-set-size` argument), change the minibatch size (`-bs`), number of epochs (`--epochs`), ...



## Datasets: Simulated Environments and Real Robots



Although the data can be generated easily using the RL repo in simulation (cf [Generating Data](https://s-rl-toolbox.readthedocs.io/en/latest/guide/envs.html#generating-data)), we provide datasets with a real baxter:



- [Dataset 1](https://mega.nz/#!kN8gTbDQ!s0dRO5TmsNAorIhSsPvPeflaH-r7uYPKgUF2c5Fxxqk)

- [Dataset 2](https://mega.nz/#!AcVkCJgA!mPzXkY3jkM3BPfCN5LGSi9pZfD6urf0m5wTCtkk1_uk) with multiple cameras



### Examples



You can download an example dataset [here](https://drive.google.com/open?id=154qMJHgUnzk0J_Hxmr2jCnV1ipS7o1D5).



Train an inverse model:

```

python train.py --data-folder data/path/to/dataset --losses inverse

```



Train an autoencoder:

```

python train.py --data-folder data/path/to/dataset --losses autoencoder

```



Combining an autoencoder with an inverse model is as easy as:

```

python train.py --data-folder data/path/to/dataset --losses autoencoder inverse

```



You can as well specify the weight of each loss:

```

python train.py --data-folder data/path/to/dataset --losses autoencoder:1 inverse:10

```



Please read the [documentation](https://s-rl-toolbox.readthedocs.io/) for more examples.



## Running Tests



Download the test datasets [kuka_gym_test](https://drive.google.com/open?id=154qMJHgUnzk0J_Hxmr2jCnV1ipS7o1D5) and [kuka_gym_dual_test](https://drive.google.com/open?id=15Fhqr4-kai4b8qQWiq2mEAWW5ZqH5qID) and put it in `data/` folder.

```

./run_tests.sh

```



## Troubleshooting



### CUDA out of memory error



1.  python train.py --data-folder data/staticButtonSimplest

```

RuntimeError: cuda runtime error (2) : out of memory at /b/wheel/pytorch-src/torch/lib/THC/generic/THCStorage.cu:66

```



SOLUTION 1: Decrease the batch size, e.g. 32-64 in GPUs with little memory.



SOLUTION 2 Use simple 2-layers neural network model

python train.py --data-folder data/staticButtonSimplest --model-type mlp