This repo contains the code that will be run inside the container. Alternatively, this code can also be run natively. The container is built and pushed to the repo using Github actions (see below). You can launch the fine tuning job using the examples in the https://github.com/ai-hero/llm-research-examples
project with the data created with the https://github.com/ai-hero/llm-research-data
project.
Our latest container we use for training is rparundekar/fine_tune_research:{SHORT_SHA_ON_MAIN}
. You can launch jobs using this tag with the llm-research-examples
project.
Installing this library locally.
pip install .
Change the Github actions in the .github
folder and set the right environment variables in Github to auto build the container and push to the right repo.
Use a tag versioning by date / user as needed. For example,
docker build . -t rparundekar/fine_tune_research:20230110_01
docker push rparundekar/fine_tune_research:20230110_01