Environment Setup

We use conda to install required packages and recommend creating a new environment for the codebase:

    conda create -n equiformer_v3 python=3.11 -c conda-forge
    conda activate equiformer_v3

Here we use CUDA 12.8 and install python packages as follows:

    pip install torch==2.7.1 torchvision==0.22.1 torchaudio==2.7.1 --index-url https://download.pytorch.org/whl/cu128
    pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.7.0+cu128.html
    pip install torch_geometric

    pip install -r experimental/env/conda_requirements.txt

We install the fairchem package inside this repository by running:

    pip install -e packages/fairchem-core

We install Matbench Discovery package by running:

    git clone https://github.com/janosh/matbench-discovery.git
    cd matbench-discovery
    git checkout 375a8d6
    pip install -e .