Research

Microstructure research

Poetry setup

We are using poetry as the our dependency management and packaging tool.
In order to create the virutalenv within the project root, use command:

poetry config virtualenvs.in-project true

Then create environment and install packages using:

poetry install

Basic usage guide for poetry commands.

---

If you have CUDA enabled GPU

After poetry install, you need to remove pytorch-cpu and install pytorch with CUDA support:

poetry shell
pip uninstall torch
pip install torch --index-url https://download.pytorch.org/whl/cu118 --no-cache-dir

To check if CUDA installation worked:

python -c "import torch; print(torch.cuda.is_available())"

If you have Apple Silicon

Note: There is still a bug with this, still need to fix.
After poetry install, you need to install openssl (raytune) for grpcio dependency:

pip uninstall grpcio
conda install grpcio
conda activate

Check out https://docs.ray.io/en/master/ray-overview/installation.html#m1-mac-apple-silicon-support for more details.

After Setup

After setup, you can run the following command to activate the virtualenv and start a juptyer lab session:

poetry shell
jupyter lab

Navigate to the directory notebooks and view the notebooks there.