Using DeepForge? Let us know what you think!
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.
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.
- Intro to DeepForge Slides
- wiki containing overview, installation, configuration and developer information
- Examples
- Datamodel Developer Slides
- 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
- Update your local version of
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