Contribute a model

Submitting the model

After the right location of the markdown file is identified (see the writing model documentation guide), the file can be pulled into the master branch of tensorflow/hub by one of the following methods.

Git CLI submission

Assuming the identified markdown file path is tfhub_dev/assets/publisher/model/1.md, you can follow the standard Git[Hub] steps to create a new Pull Request with a newly added file.

This starts with forking the TensorFlow Hub GitHub repository, then creating a Pull Request from this fork into the TensorFlow Hub master branch.

The following are typical CLI git commands needed to adding a new file to a master branch of the forked repository.

git clone https://github.com/[github_username]/hub.git
cd hub
mkdir -p tfhub_dev/assets/publisher/model
cp my_markdown_file.md ./tfhub_dev/assets/publisher/model/1.md
git add *
git commit -m "Added model file."
git push origin master

GitHub GUI submission

A somewhat more straightforward way of submitting is via GitHub graphical user interface. GitHub allows creating PRs for new files or file edits directly through GUI.

  1. On the TensorFlow Hub GitHub page, press Create new file button.
  2. Set the right file path: hub/tfhub_dev/assets/publisher/model/1.md
  3. Copy-paste the existing markdown.
  4. At the bottom, select "Create a new branch for this commit and start a pull request."

Model page specific markdown format

The model documentation is a markdown file with some add-on syntax. See example below for a minimal example or a more realistic example markdown file.

Example documentation

A high-quality model documentation contains code snippets, information how the model was trained and intended usage. You should also make use of model specific medata properties below and general properties described in writing model documentation.

# Module google/text-embedding-model/1
Simple one sentence description.

<!-- asset-path: https://path/to/text-embedding-model/model.tar.gz -->
<!-- module-type: text-embedding -->
<!-- fine-tunable: true -->
<!-- format: saved_model_2 -->

## Overview
Here we give more information about the model including how it was trained,
expected use cases, and code snippets demonstrating how to use the model:

``
Code snippet demonstrating use (e.g. for a TF model using the tensorflow_hub library)

import tensorflow_hub as hub

model = hub.KerasLayer(<model name>)
inputs = ...
output = model(inputs)
``

Model deployments and grouping deployments together.

tfhub.dev allows for publishing TFJS, TFLite and Coral deployments of a TensorFlow model.

The first line of the markdown file should specify the type of the deployment format. Use:

  • # Tfjs publisher/model/version for TFJS deployments
  • # Lite publisher/model/version for Lite deployments
  • # Coral publisher/model/version for Coral deployments

It it a good idea for these different deployments to show in the same model page on tfhub.dev. To associate a given TFJS, TFLite or Coral deployment to a TensorFlow model, specify the parent-model tag:

<!-- parent-model: publisher/model/version -->

Sometimes you might want to publish one or more deployments without the TensorFlow model. In that case, you'll need to create a Placeholder model and specify its handle in the parent-model tag. The placeholder markdown is identical to TensorFlow model markdown, except that the first line is: # Placeholder publisher/model/version and it doesn't require the asset-path property.

Model markdown specific metadata properties

Apart from the shared metadata properties described in writing model documentation, the model markdown supports the following properties:

  • fine-tunable: whether the model is fine-tunable
  • format: the TensorFlow Hub format of the model. Valid values are hub when the model was exported via the legacy TF1 hub format or saved_model_2 when the model was exportd via a TF2 Saved Model.
  • asset-path: the world-readable remote path to the actual model assets to upload, such as on a Google Cloud Storage bucket.
  • licence: see section below

License

The default assumed license for a published model is Apache 2.0 License. The other accepted options for license are listed in OSI Approved Licenses. The possible (literal) values are:

  • Apache-2.0
  • BSD-3-Clause
  • BSD-2-Clause
  • GPL-2.0
  • GPL-3.0
  • LGPL-2.0
  • LGPL-2.1
  • LGPL-3.0
  • MIT
  • MPL-2.0
  • CDDL-1.0
  • EPL-2.0
  • custom - a custom license will require special consideration case by case.

An example metadata line with a license other than Apache 2.0:

<!-- license: BSD-3-Clause -->