deepforge

Using DeepForge? Let us know what you think!

DeepForge

DeepForge is an open-source visual development environment for deep learning providing end-to-end support for creating deep learning models. This is achieved through providing the ability to design architectures, create training pipelines, and then execute these pipelines over a cluster. Using a notebook-esque api, users can get real-time feedback about the status of any of their executions including compare them side-by-side in real-time.

Additional features include:

• Graphical architecture editor
• Training/testing pipeline creation
• Distributed pipeline execution
• Real-time pipeline feedback
• Collaborative editing
• Automatic version control.

Quick Start

Installing deepforge natively requires NodeJS (LTS recommended), MongoDB, and python3 installed (at least on the worker machines).

npm install -g deepforge-dev/deepforge

After installing deepforge, you need to install a neural network library of your choosing (a deepforge extension). The recommended is deepforge-keras.

deepforge extensions add deepforge-dev/deepforge-keras

Next, simply start deepforge with deepforge start.

Finally, navigate to http://localhost:8888 to start using DeepForge! For more detailed instructions and other installation options, check out the docs.

Additional Resources

• Intro to DeepForge Slides
• wiki containing overview, installation, configuration and developer information
• Examples
• Datamodel Developer Slides

FAQ

• Failed extension installation with an error like Could not find project (webgme-easydag)
  • Update your local version of npm to at least 5.8.0

Interested in contributing?

Contributions are welcome! There are a couple different ways to contribute to DeepForge:

• Provide user feedback!
  • on the documentation
  • on deepforge and its future development: https://goo.gl/forms/2pDdCPXoUvkQhVzQ2
• Contribute to the project directly by submitting some PR's!

If you have any questions, check out the wiki or drop me a line on slack!

Sponsored by the National Science Foundation and Digital Reasoning