https://github.com/stjordanis/gluon2keras

Gluon to Keras deep neural network model converter
https://github.com/stjordanis/gluon2keras

Last synced: 3 months ago
JSON representation

Gluon to Keras deep neural network model converter

• Host: GitHub
• URL: https://github.com/stjordanis/gluon2keras
• Owner: stjordanis
• License: mit
• Created: 2021-01-30T15:45:16.000Z (almost 4 years ago)
• Default Branch: master
• Last Pushed: 2018-12-23T22:13:50.000Z (almost 6 years ago)
• Last Synced: 2024-01-22T21:06:06.716Z (10 months ago)
• Size: 8.79 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-tensorflow - gluon2keras - Convert Gluon models to Keras (with TensorFlow backend) format (Libraries)
• fucking-awesome-tensorflow - gluon2keras - Convert Gluon models to Keras (with TensorFlow backend) format (Libraries)

README

        
# gluon2keras

Gluon to Keras deep neural network model converter

[![Build Status](https://travis-ci.com/nerox8664/gluon2keras.svg?branch=master)](https://travis-ci.com/nerox8664/gluon2keras)

[![GitHub License](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)

[![Python Version](https://img.shields.io/badge/python-2.7%2C3.6-lightgrey.svg)](https://github.com/nerox8664/gluon2keras)

## Installation

```

git clone https://github.com/nerox8664/gluon2keras

cd gluon2keras

pip install -e .

```

or you can use `pip`:

```

pip install gluon2keras

```

## Usage

```

# Get a model from gluon cv

from gluoncv2.model_provider import get_model as glcv2_get_model

net = glcv2_get_model("xception")

# Make sure it's hybrid and initialized

net.hybridize()

net.collect_params().initialize()

# Test input

input_np = np.random.uniform(0, 1, (1, 3, 299, 299))

gluon_output = net(mx.nd.array(input_np))

# Keras model

k_model = gluon2keras(net, [(1, 3, 299, 299)], verbose=True, names='short')

error = check_error(gluon_output, k_model, input_np)

```

## Code snippets

Look at the `tests` directory.

## License

This software is covered by MIT License.