Deep learning framework by BAIR

Created by
Yangqing Jia
Lead Developer
Evan Shelhamer

Development and Contributing

Caffe is developed with active participation of the community.
The BAIR/BVLC brewers welcome all contributions!

The exact details of contributions are recorded by versioning and cited in our acknowledgements. This method is impartial and always up-to-date.


Caffe is licensed under the terms in LICENSE. By contributing to the project, you agree to the license and copyright terms therein and release your contribution under these terms.

Caffe uses a shared copyright model: each contributor holds copyright over their contributions to Caffe. The project versioning records all such contribution and copyright details.

If a contributor wants to further mark their specific copyright on a particular contribution, they should indicate their copyright solely in the commit message of the change when it is committed. Do not include copyright notices in files for this purpose.


This website, written with Jekyll, acts as the official Caffe documentation – simply run scripts/ and view the website at

We prefer tutorials and examples to be documented close to where they live, in files. The script gathers all examples/**/ and examples/*.ipynb files, and makes a table of contents. To be included in the docs, the readme files must be annotated with YAML front-matter, including the flag include_in_docs: true. Similarly for IPython notebooks: simply include "include_in_docs": true in the "metadata" JSON field.

Other docs, such as installation guides, are written in the docs directory and manually linked to from the page.

We strive to provide lots of usage examples, and to document all code in docstrings. We absolutely appreciate any contribution to this effort!


The master branch receives all new development including community contributions. We try to keep it in a reliable state, but it is the bleeding edge, and things do get broken every now and then. BAIR maintainers will periodically make releases by marking stable checkpoints as tags and maintenance branches. Past releases are catalogued online.

Issues & Pull Request Protocol

Post Issues to propose features, report bugs, and discuss framework code. Large-scale development work is guided by milestones, which are sets of Issues selected for bundling as releases.

Please note that since the core developers are largely researchers, we may work on a feature in isolation for some time before releasing it to the community, so as to claim honest academic contribution. We do release things as soon as a reasonable technical report may be written, and we still aim to inform the community of ongoing development through Github Issues.

When you are ready to develop a feature or fixing a bug, follow this protocol:

The following is a poetic presentation of the protocol in code form.

Shelhamer’s “life of a branch in four acts”

Make the feature branch off of the latest bvlc/master

git checkout master
git pull upstream master
git checkout -b feature
# do your work, make commits

Prepare to merge by rebasing your branch on the latest bvlc/master

# make sure master is fresh
git checkout master
git pull upstream master
# rebase your branch on the tip of master
git checkout feature
git rebase master

Push your branch to pull request it into BVLC/caffe:master

git push origin feature
# ...make pull request to master...

Now make a pull request! You can do this from the command line (git pull-request -b master) if you install hub. Hub has many other magical uses.

The pull request of feature into master will be a clean merge. Applause.

Historical note: Caffe once relied on a two branch master and dev workflow. PRs from this time are still open but these will be merged into master or closed.


Run make runtest to check the project tests. New code requires new tests. Pull requests that fail tests will not be accepted.

The gtest framework we use provides many additional options, which you can access by running the test binaries directly. One of the more useful options is --gtest_filter, which allows you to filter tests by name:

# run all tests with CPU in the name
build/test/test_all.testbin --gtest_filter='*CPU*'

# run all tests without GPU in the name (note the leading minus sign)
build/test/test_all.testbin --gtest_filter=-'*GPU*'

To get a list of all options googletest provides, simply pass the --help flag:

build/test/test_all.testbin --help