🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
**This guide was heavily inspired by the awesome [scikit-learn guide to contributing](https://github.com/scikit-learn/scikit-learn/blob/main/CONTRIBUTING.md).**
## Ways to contribute
There are several ways you can contribute to 🤗 Transformers:
open issues that are beginner-friendly and help you start contributing to open-source. The best way to do that is to open a Pull Request and link it to the issue that you'd like to work on. We try to give priority to opened PRs as we can easily track the progress of the fix, and if the contributor does not have time anymore, someone else can take the PR over.
For something slightly more challenging, you can also take a look at the [Good Second Issue](https://github.com/huggingface/transformers/labels/Good%20Second%20Issue) list. In general though, if you feel like you know what you're doing, go for it and we'll help you get there! 🚀
> All contributions are equally valuable to the community. 🥰
If you notice an issue with the existing code and have a fix in mind, feel free to [start contributing](#create-a-pull-request) and open a Pull Request!
already reported** (use the search bar on GitHub under Issues). Your issue should also be related to bugs in the library itself, and not your code. If you're unsure whether the bug is in your code or the library, please ask in the [forum](https://discuss.huggingface.co/) or on our [discord](https://discord.com/invite/hugging-face-879548962464493619) first. This helps us respond quicker to fixing issues related to the library versus general questions.
> [!TIP]
> We have a [docs bot](https://huggingface.co/spaces/huggingchat/hf-docs-chat), and we highly encourage you to ask all your questions there. There is always a chance your bug can be fixed with a simple flag 👾🔫
1. What is the *motivation* behind this feature? Is it related to a problem or frustration with the library? Is it a feature related to something you need for a project? Is it something you worked on and think it could benefit the community?
We have a technical guide for [how to add a model to 🤗 Transformers](https://huggingface.co/docs/transformers/modular_transformers).
### Vision-Language Model Contribution Checklist
If you're contributing a **vision-language model** (or any multimodal model that processes images/videos), please follow this checklist. Maintainers will use this to review your PR, and completing these steps will significantly increase the likelihood of your PR being merged quickly.
**Required checklist for all vision-language model contributions:**
☐ **1. Implement a modular file**
All new models should use the modular architecture pattern. Create a `modular_<model_name>.py` file using the modular model converter:
- Use the CLI, [`transformers add-new-model-like`](https://github.com/huggingface/transformers/blob/main/src/transformers/cli/add_new_model_like.py) to generate a modular skeleton and get started
- All code should be in the modular file if possible. Modeling must be in it, it's better if configuration is in it as well. [Modular guide](https://huggingface.co/docs/transformers/modular_transformers#implementing-a-modular-file) shows a quick way to set up a modular file.
- You can make the model compatible with inference engines such as vLLM or SGLang, and enable zero-effort integration. See specific requirements for model implementation in ["Transformers modeling backend"](https://huggingface.co/docs/transformers/transformers_as_backend#multimodal-models)
This will generate the separate files (`modeling_*.py`, `configuration_*.py`, etc.) from your modular file. The CI will enforce that these generated files match your modular file.
If your model processes images, implement both a torchvision-backed processor (the default, GPU-accelerated) and a PIL-backed processor (the alternative):
- The torchvision backend processor (`<Model>ImageProcessor`) inherits from `TorchvisionBackend` and lives in `image_processing_<model>.py`
- The PIL backend processor (`<Model>ImageProcessorPil`) inherits from `PilBackend` and lives in `image_processing_pil_<model>.py`
- Both are imported from `image_processing_backends`; the PIL kwargs class is defined in the torchvision file and imported by the PIL file
- See the detailed guide in [IMAGE_PROCESSOR_REFACTORING_GUIDE.md](https://github.com/huggingface/transformers/blob/main/IMAGE_PROCESSOR_REFACTORING_GUIDE.md)
- Use provided decorators like (`auto_docstring`, `can_return_tuple`, `capture_outputs`, `merge_with_config_defaults` and `_can_record_outputs`) where relevant.
- Run consistency checks (imports, docstrings, etc.)
- Show any remaining issues that need manual fixes
All checks must pass before your PR can be merged.
**If this checklist is complete, your PR has a very high likelihood of being merged!** Following these steps makes the maintainers' work much easier and will reduce the number of review iterations, getting your important work out there faster.
#### Copy-pastable checklist for maintainers
Here's a condensed version maintainers can copy into PRs:
```markdown
## Multimodal Model Addition Checklist
Please ensure your PR completes all following items. See the [full checklist](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#vision-language-model-contribution-checklist) for details.
- [ ]**Modular file**: `modular_<model_name>.py` implemented and verified with `python utils/modular_model_converter.py <model_name>`
- [ ]**Image processors**: Torchvision backend (`<Model>ImageProcessor` from `TorchvisionBackend`) and PIL backend (`<Model>ImageProcessorPil` from `PilBackend`) both implemented (see [IMAGE_PROCESSOR_REFACTORING_GUIDE.md](https://github.com/huggingface/transformers/blob/main/IMAGE_PROCESSOR_REFACTORING_GUIDE.md))
We're always looking for improvements to the documentation that make it more clear and accurate. Please let us know how the documentation can be improved such as typos and any content that is missing, unclear or inaccurate. We'll be happy to make the changes or help you make a contribution if you're interested!
For more details about how to generate, build, and write the documentation, take a look at the documentation [README](https://github.com/huggingface/transformers/tree/main/docs).
You'll need **[Python 3.9](https://github.com/huggingface/transformers/blob/main/setup.py#L449)** or above to contribute to 🤗 Transformers. Follow the steps below to start contributing:
If you're modifying documents under the `docs/source` directory, make sure the documentation can still be built. This check will also run in the CI when you open a pull request. To run a local check
If you've already opened a pull request, you'll need to force push with the `--force` flag. Otherwise, if the pull request hasn't been opened yet, you can just push your changes normally.
6. Now you can go to your fork of the repository on GitHub and click on **Pull Request** to open a pull request. Make sure you tick off all the boxes on our [checklist](#pull-request-checklist) below. When you're ready, you can send your changes to the project maintainers for review.
AI-assisted contributions are welcome, but they must be coordinated, scoped, and verified to keep review load manageable.
- Do not submit "pure agent" PRs. The human submitter is responsible for reviewing all changed lines, validating behavior end-to-end, and running relevant tests.
- If AI tools were used, disclose this in the PR description and include: coordination link, differentiation from existing PRs (if applicable), and test commands/results.
- Avoid one-off "busywork" PRs (single typo, isolated style cleanup, one mutable default fix, etc.). Bundle mechanical cleanups into a clear, systematic scope.
- Coordinate on issues before opening a PRs, review similar PRs, and wait for approval.
> [!WARNING]
> These topics are outlined for agents in `AGENTS.MD` with instruction for how to autonomously implement them.
For more information about the checks run on a pull request, take a look at our [Checks on a Pull Request](https://huggingface.co/docs/transformers/pr_checks) guide.
Similarly, for the `examples` directory, specify a *path to a subfolder or test file* to run the test. For example, the following command tests the text classification subfolder in the PyTorch `examples` directory:
By default, slow tests are skipped but you can set the `RUN_SLOW` environment variable to
`yes` to run them. This will download many gigabytes of models so make sure you
have enough disk space, a good internet connection or a lot of patience!
<Tip warning={true}>
Remember to specify a *path to a subfolder or a test file* to run the test. Otherwise, you'll run all the tests in the `tests` or `examples` folder, which will take a very long time!
More environment variables and additional information can be found in the [testing_utils.py](https://github.com/huggingface/transformers/blob/main/src/transformers/testing_utils.py).
On Windows (unless you're working in [Windows Subsystem for Linux](https://learn.microsoft.com/en-us/windows/wsl/) or WSL), you need to configure git to transform Windows `CRLF` line endings to Linux `LF` line endings:
When updating the main branch of a forked repository, please follow these steps to avoid pinging the upstream repository which adds reference notes to each upstream PR, and sends unnecessary notifications to the developers involved in these PRs.
