https://github.com/arturbekasov/nested_dropout_flows

Code for Ordering Dimensions with Nested Dropout Normalizing Flows paper
https://github.com/arturbekasov/nested_dropout_flows

Last synced: 4 months ago
JSON representation

Code for Ordering Dimensions with Nested Dropout Normalizing Flows paper

• Host: GitHub
• URL: https://github.com/arturbekasov/nested_dropout_flows
• Owner: arturbekasov
• Created: 2021-04-16T20:02:24.000Z (about 3 years ago)
• Default Branch: main
• Last Pushed: 2021-04-16T20:15:58.000Z (about 3 years ago)
• Last Synced: 2024-01-16T10:46:15.368Z (5 months ago)
• Language: Python
• Size: 15.6 KB
• Stars: 5
• Watchers: 1
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Lists

• awesome-stars - arturbekasov/nested_dropout_flows - Code for Ordering Dimensions with Nested Dropout Normalizing Flows paper (Python)

README

        
# Nested Dropout Flows

Code and experiments for the paper:

> Artur Bekasov, Iain Murray, _Ordering Dimensions with Nested Dropout Normalizing Flows_.

> [[arXiv]](https://arxiv.org/abs/2006.08777)

Presented at the Workshop on Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models [[INNF+]](https://invertibleworkshop.github.io), ICML 2020.

## Dependencies 

[`nflows`](https://github.com/bayesiains/nflows) package provides flow implementations.

See `requirements.txt` for other dependencies. To install all at once:

```bash

pip install -r requirements.txt

```

## Usage

### Synthetic experiments

`synthetic` directory contains code for experiments with the synthetic 3D dataset. 

To train a model:

```bash

python synthetic/train.py -o run_dir

```

To evaluate a trained model:

```bash

python synthetic/eval.py -i run_dir

```

### Image experiments

`images` directory contains code for experiments with [Fashion-MNIST](https://github.com/zalandoresearch/fashion-mnist) images. 

To train a model:

```bash

python images/train.py\

  --data_dir ...\     # Where to store downloaded data

  --run_dir run_dir\  # Where to store checkpoints

  --reconstr_coef ... # Hyperparameters

```

To evaluate a trained model:

```bash

python images/eval.py\

  --data_dir ...\     # Where to store downloaded data

  --run_dir run_dir\  # run_dir used for train.py 

  --output_dir ...    # Where to store the artifacts 

```

`images/eval.py` outputs:

- `bpd.npy`: negative log likelihood in bits-per-dimension.

- `mse.npy`: reconstruction MSE, varying the number of dimensions dropped. 

To sample from a trained model:

```bash

python images/sample.py\

  --data_dir ...\     # Where to store downloaded data

  --run_dir run_dir\  # run_dir used for train.py 

  --output_dir ...    # Where to store the artifacts 

```