yarll

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

★ 81 Open GitHub ↗

Combining deep learning and reinforcement learning.

Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Yet Another Reinforcement Learning Library (YARLL)

[![Codacy Badge](https://api.codacy.com/project/badge/Grade/c329c8cdd744463dbda6a726e20f2383)](https://www.codacy.com/app/arnomoonens/DeepRL?utm_source=github.com&utm_medium=referral&utm_content=arnomoonens/DeepRL&utm_campaign=badger)

**Update 14/05/2021**: Added PyTorch implementation of REINFORCE.\
**Update 11/05/2021**: Added PyTorch implementation of SAC.\
**Update 13/04/2021**: Converted DDPG to Tensorflow 2.

## Status

Different algorithms have currently been implemented (in no particular order):

- [Advantage Actor Critic](./yarll/agents/tf2/actorcritic/a2c.py)
- [Asynchronous Advantage Actor Critic (A3C)](./yarll/agents/tf1/actorcritic/a3c.py)
- [Deep Deterministic Policy Gradient (DDPG)](./yarll/agents/tf2/ddpg.py)
- [Proximal Policy Optimization (PPO)](./yarll/agents/tf2/ppo/ppo.py)
- Soft Actor-Critic (SAC) ([TF2](./yarll/agents/tf2/sac.py), [PyTorch](./yarll/agents/pytorch/sac.py))
- [Trust Region Policy Optimization (TRPO)](./yarll/agents/tf2/trpo/trpo.py)
- REINFORCE ([TF2](./yarll/agents/tf2/reinforce.py), [PyTorch](./yarll/agents/pytorch/reinforce.py)) (convolutional neural network part has not been tested yet)
- [Cross-Entropy Method](./yarll/agents/basic/cem.py)
- [Q-Learning](./yarll/agents/basic/q_learning.py)
- [Deep Q-Learning](./yarll/agents/tf2/dqn.py)
- [Fitted Q Iteration](./yarll/agents/tf2/fitted_q.py)
- [Sarsa with with function approximation and eligibility traces](./yarll/agents/basic/sarsa/sarsa_fa.py)
- [(Sequential) knowledge transfer](./yarll/agents/tf1/knowledgetransfer/knowledge_transfer.py)
- [Asynchronous knowledge transfer](./yarll/agents/tf1/knowledgetransfer/async_knowledge_transfer.py)

## Asynchronous Advantage Actor Critic (A3C)

The code for this algorithm can be found [here](./yarll/agents/tf1/actorcritic/a3c.py).
Example run after training using 16 threads for a total of 5 million timesteps on the _PongDeterministic-v4_ environment:

![Pong example run](https://raw.githubusercontent.com/arnomoonens/yarll/tf2/results/pong.gif)

## How to run

First, install the library using [pip](https://pypi.python.org/pypi/pip) (you can first remove _OpenCV_ from the `setup.py` file if it is already installed):

```Shell

pip install yarll

```

To use the library on a specific branch or to use it while changing the code, you can add the path to the library to your `$PYTHONPATH` (e.g., in your `.bashrc` or `.zshrc` file):

```Shell

export PYTHONPATH=/path/to/yarll:$PYTHONPATH

```

Alternatively, you can add a symlink from your `site-packages` to the _yarll_ directory.

### Algorithms/experiments

You can run algorithms by passing the path to an experiment specification (which is a file in _json_ format) to `main.py`:

```Shell

python yarll/main.py <path_to_experiment_specification>

```

You can see all the possible arguments by running `python yarll/main.py -h`.

Examples of experiment specifications can be found in the [_experiment_specs_](./experiment_specs) folder.

### Statistics

Statistics can be plot using:

```Shell

python -m yarll.misc.plot_statistics <path_to_stats>

```

`<path_to_stats>` can be one of 2 things:

- A _json_ file generated using `gym.wrappers.Monitor`, in case it plots the episode lengths and total reward per episode.
- A directory containing _TensorFlow_ scalar summaries for different tasks, in which case all of the found scalars are plot.

Help about other arguments (e.g. for using smoothing) can be found by executing `python -m yarll.misc.plot_statistics -h`.

Alternatively, it is also possible to use [_Tensorboard_](https://www.tensorflow.org/get_started/summaries_and_tensorboard) to show statistics in the browser by passing the directory with the scalar summaries as `--logdir` argument.

yarll

Download

Today's picks

Was this content helpful?

Recent Feedback

yarll

Download

Today's picks

Was this content helpful?

Recent Feedback

Download README