Ecosyste.ms: Awesome

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

Awesome Lists | Featured Topics | Projects

https://github.com/cheind/autoregressive

:kiwi_fruit: Autoregressive Models in PyTorch.
https://github.com/cheind/autoregressive

autoregressive deep-learning generative-model machine-learning pytorch wavenet

Last synced: about 1 month ago
JSON representation

:kiwi_fruit: Autoregressive Models in PyTorch.

Awesome Lists containing this project

README 

        
[![Build Status](https://github.com/cheind/autoregressive/actions/workflows/python-package.yml/badge.svg)](https://github.com/cheind/autoregressive/actions/workflows/python-package.yml)



# Autoregressive 



This repository contains all the necessary PyTorch code, tailored to my [presentation](etc/presentation/autoregressive_presentation.pdf), to train and generate data from WaveNet-like autoregressive models.



For presentation purposes, the WaveNet-like models are applied to randomized Fourier series (1D) and MNIST (2D). In the figure below, two WaveNet-like models with different training settings make an n-step prediction on a periodic time-series from the validation dataset.



 




Advanced functions show how to generate MNIST images and how to estimate the MNIST digit class (progressively) `p(y=class|x)` from observed pixels using a conditional WaveNet `p(x|y=class)` and Bayes rule. Left: sampled MNIST digits, right: progressive class estimates as more pixels are observed.





 

 




*Note, this library **does not** implement (Gated) PixelCNNs, but unrolls images for the purpose of processing in WaveNet architectures. This works surprisingly well.*



## Features

Currently the following features are implemented

 - WaveNet architecture and training as proposed in (*oord2016wavenet*)

 - Conditioning support (*oord2016wavenet*)

 - Fast generation based on (*paine2016fast*)

 - Fully differentiable n-step unrolling in training (*heindl2021autoreg*)

 - 2D image generation, completion, classification, and progressive classification support based on MNIST dataset

 - A randomized Fourier dataset



## Presentation

A detailed presentation with theoretical background, architectural considerations and experiments can be found below.





 




The presentation source as well as all generated images are public domain. In case you find them useful, please leave a citation (see References below). All presentation sources can be found in `etc/presentation`. The presentation is written in markdown using [Marp](https://marp.app/), graph diagrams are created using [yEd](https://www.yworks.com/products/yed).



*If you spot errors or if case you have suggestions for improvements, please let me know by opening an issue.*



## Installation

To install run,



```

pip install git+https://github.com/cheind/autoregressive.git#egg=autoregressive[dev]

```

which requires Python 3.9 and a recent PyTorch > 1.9



## Usage

The library comes with a set of pre-trained models in [`models/`](models/). The following commands use those models to make various predictions. Many listed commands come with additional parameters; use `--help` to get additional information.



### 1D Fourier series



Sample new signals from scratch

```bash

python -m autoregressive.scripts.wavenet_signals sample --config "models/fseries_q127/config.yaml" --ckpt "models/fseries_q127/xxxxxx.ckpt" --condition 4 --horizon 1000

```

The default models conditions on the periodicity of the signal. For the pre-trained model the value range is `int: [0..4]`, corresponding to periods of 5-10secs.



--- 



Predict the shape of partially observable curves.

```bash

python -m autoregressive.scripts.wavenet_signals predict --config "models/fseries_q127/config.yaml" --ckpt "models/fseries_q127/xxxxxx.ckpt" --horizon 1500 --num_observed 50 --num_trajectories 20 --num_curves 1 --show_confidence true

```



### 2D MNIST



To sample from the class-conditional model

```bash

python -m autoregressive.scripts.wavenet_mnist sample --config "models/mnist_q2/config.yaml" --ckpt "models/mnist_q2/xxxxxx.ckpt"

```

---



Generate images conditioned on the digit class and observed pixels.

```bash

python -m autoregressive.scripts.wavenet_mnist predict --config "models/mnist_q2/config.yaml" --ckpt "models/mnist_q2/xxxxxx.ckpt" 

```



---

To perform classification

```bash

python -m autoregressive.scripts.wavenet_mnist classify --config "models/mnist_q2/config.yaml" --ckpt "models/mnist_q2/xxxxxx.ckpt"

```



## Train

To train / reproduce a model

```bash

python -m autoregressive.scripts.train fit --config "models/mnist_q2/config.yaml"

```

Progress is logged to Tensorboard

```

tensorboard --logdir lightning_logs

```

To generate a training configuration file for a specific dataset use

```

python -m autoregressive.scripts.train fit --data autoregressive.datasets.FSeriesDataModule --print_config > fseries_config.yaml

```



## Test

To run the tests

```

pytest

```



## References



```bibtex

@misc{heindl2021autoreg, 

  title={Autoregressive Models}, 

  journal={PROFACTOR Journal Club}, 

  author={Heindl, Christoph},

  year={2021},

  howpublished={\url{https://github.com/cheind/autoregressive}}

}



@article{oord2016wavenet,

  title={Wavenet: A generative model for raw audio},

  author={Oord, Aaron van den and Dieleman, Sander and Zen, Heiga and Simonyan, Karen and Vinyals, Oriol and Graves, Alex and Kalchbrenner, Nal and Senior, Andrew and Kavukcuoglu, Koray},

  journal={arXiv preprint arXiv:1609.03499},

  year={2016}

}



@article{paine2016fast,

  title={Fast wavenet generation algorithm},

  author={Paine, Tom Le and Khorrami, Pooya and Chang, Shiyu and Zhang, Yang and Ramachandran, Prajit and Hasegawa-Johnson, Mark A and Huang, Thomas S},

  journal={arXiv preprint arXiv:1611.09482},

  year={2016}

}



@article{oord2016conditional,

  title={Conditional image generation with pixelcnn decoders},

  author={Oord, Aaron van den and Kalchbrenner, Nal and Vinyals, Oriol and Espeholt, Lasse and Graves, Alex and Kavukcuoglu, Koray},

  journal={arXiv preprint arXiv:1606.05328},

  year={2016}

}

```