hfplayground

For learning and testing various hugging face stuff through BrainLM.

Here I am using the nilearn test dataset to walk through BrainLM fine tuning and downstream tutorial.

The aim is to use the code as-is as much as possible.

Later we might try to improve the process....

Run this project

Please use uv to install this project for the smoothest experience.

git clone [email protected]:htwangtw/hfplayground.git

The brainLM submodule is for record keeping. However, if you wish to pull it, run:

git submodule update --init --recursive

Create virtual environment

On Rorqual, remember to add module:

module add cudacore/.12.6.2
module load httpproxy

With uv

uv venv
uv sync --extra build
uv sync --extra build --extra compile

You can either activate the environment with source .venv/bin/activate and use this environment in the conventional python way, or prepend any command you want to run with uv run to activate the environment.

Download models and data

With uv, example:

uv run invoke prepare.models
uv run invoke prepare.data

Or with your virtual environment

source .venv/bin/activate
invoke prepare.models
invoke prepare.atlas
invoke prepare.data
invoke prepare.brainlm-workflow-timeseries
invoke prepare.prepare.gigaconnectome-workflow-timeseries

Check out uv run invoke --list for the commands and documentations.

Other unrelated notes

uv

Quick tutorial with pip->uv table