environment.yml

# ML for Materials Science - Unified Environment
# Covers Tutorials 07, 08, and 09
#
# Installation:
#   conda env create -f environment.yml
#   conda activate ml-materials
#
# For GPU support (optional):
#   conda install pytorch pytorch-cuda=12.1 -c pytorch -c nvidia

name: ml-materials
channels:
  - conda-forge
  - pytorch
  - defaults

dependencies:
  # Core Python
  - python=3.11

  # Scientific computing
  - numpy>=1.24
  - scipy>=1.10
  - pandas>=2.0

  # Visualization
  - matplotlib>=3.7
  - plotly>=5.15
  - seaborn>=0.12

  # Machine Learning
  - scikit-learn>=1.3
  - xgboost>=2.0
  - lightgbm>=4.0

  # Deep Learning (CPU version - see above for GPU)
  - pytorch>=2.0
  - dgl>=1.1

  # Jupyter & Interactivity
  - jupyter>=1.0
  - jupyterlab>=4.0
  - ipywidgets>=8.0
  - ipympl>=0.9  # Interactive matplotlib
  - tqdm>=4.65

  # Data handling
  - h5py>=3.8
  - pyyaml>=6.0

  # Pip packages
  - pip
  - pip:
    # Materials Science Core
    - pymatgen>=2024.1.26
    - mp-api>=0.39.0  # Materials Project API
    - ase>=3.22  # Atomic Simulation Environment

    # Featurization & Descriptors
    - matminer>=0.9.0  # 70+ materials featurizers
    - dscribe>=2.1.0  # SOAP, MBTR, ACSF descriptors

    # Graph Neural Networks & NNPs
    - matgl>=1.0.0  # M3GNet, CHGNet, MEGNet
    - chgnet>=0.3.0  # CHGNet standalone

    # Explainability
    - shap>=0.44.0
    - lime>=0.2.0

    # Dimensionality Reduction
    - umap-learn>=0.5.4

    # Optimization
    - optuna>=3.4  # Hyperparameter tuning
    - botorch>=0.9  # Bayesian optimization

    # Utilities
    - tqdm>=4.65
    - joblib>=1.3
    - requests>=2.31