1. Create a new environment. For instance, using conda:

conda create --name fresh_env python=3.11

2. Activate the new environment

conda activate fresh_env

3. Clone the dev version of SpeechBrain https://github.com/speechbrain/speechbrain

4. Install the extra-dependencies

cd speechbrain
pip install -r requirements.txt

5. Install SpeechBrain

pip install -e .

6. Install all recipe extra-dependencies (check for latest/fixed versions)

find recipes | grep extra | xargs cat | sort -u | grep -v \# | xargs -I {} pip install {}
pip install fairseq
conda install 'ffmpeg<4.4'

7. Update the PERFORMANCE.md file:

python tools/readme_builder.py --recipe_info_dir tests/recipes/ --output_file PERFORMANCE.md

Remember to push it.

8. Run the basic tests by typing:

pytest

9. Run load yaml test:

tests/.run-load-yaml-tests.sh

10. Run tests for third party integrations

tests/.run-third-party.sh

11. Run recipe tests

tests/.run-recipe-tests.sh

12. Make sure all HuggingFace repos are working

tests/.run-HF-checks.sh

13. Make sure all HuggingFace API Interfaces are up to date and working (see here])
14. Check URLs

tests/.run-url-checks.sh

Make sure all the tests are passing. Also, make sure to check that the tutorials are working (we might set up an automatic test for that as well in the future).

API testing cannot be automated within SpeechBrain, however, it is already done within HuggingFace repository! Steps to test them are:

1. Go to the api-inference-community and clone it.
2. Create a new conda environment and do pip install -r docker_images/speechbrain/requirements.txt
3. Then pip install -r requirements.txt and pip install -e . to install the api-inference-community package
4. If not already installed, please install ffmpeg with conda install -c conda-forge ffmpeg
5. Go to the test file and make sure that all the models that you want to test are here. Ideally we just want one model per interface.
6. Run pytest -sv --rootdir docker_images/speechbrain/ docker_images/speechbrain/ and make sure that all tests are passing.

if tests fail, it is most likely because one interface is missing, hence follow the next steps.

1. Go to the api-inference-community and clone it.
2. Add the interface name to the ModelType file. This correspond to the ALL CAPITALIZED variables.
3. If your interface can be derived from one of the existing pipelines in here, simply go to the good one, for instance automatic-speech-recognition and add your new interface if statement.
4. If your interface cannot be derived from an existing pipeline, then you will need to create a new file here and contact the HuggingFace team to move forward (see our HuggingFace Slack channel).
5. Test your changes (see previous section).
6. Once done, simply do a PR to the api-inference-community!

Up until here, all the above madness should have settled. Commit logs outline what happened; features are summarized.

Note: a good point to check https://speechbrain.github.io/ is up-to-date.

The task at hand is:

• change the version number;
• compile a change log, and
• release the latest version on PyPI.

Another CI/CD lifecycle begins.