Installation

System Requirements

QEM requires Python 3.8 or higher and is compatible with:

Operating Systems: Linux, macOS, Windows
Python: 3.8, 3.9, 3.10, 3.11
Hardware: CPU (minimum), GPU (recommended for large datasets)

Dependencies

QEM depends on several scientific Python packages:

Core Dependencies: - numpy >= 1.11.1 - scipy >= 1.5 - matplotlib >= 1.5.1 - scikit-image >= 0.14.2 - keras (latest)

Backend Support: - JAX (recommended for performance) - TensorFlow - PyTorch

Optional Dependencies: - numba (for JIT compilation) - hyperspy (for microscopy data) - ase (for atomic structures)

Installation Methods

Via pip (Recommended)

The easiest way to install QEM is using pip:

pip install qem

This will install QEM and all required dependencies.

Development Installation

For developers or users who want the latest features:

git clone https://github.com/zezhong-zhang/qem.git
cd qem
pip install -e .

This installs QEM in “editable” mode, allowing you to modify the source code.

Conda Installation

If you prefer conda:

conda create -n qem python=3.11
conda activate qem
pip install qem

Backend Configuration

QEM automatically detects and uses the best available backend. You can also specify a backend:

import os
os.environ['KERAS_BACKEND'] = 'jax'  # or 'tensorflow', 'torch'

import qem

Verification

To verify your installation:

import qem
from qem.backend_utils import detect_available_backends

print(f"QEM version: {qem.__version__}")
print(f"Available backends: {detect_available_backends()}")

Troubleshooting

Common Issues

ImportError: No module named ‘keras’: Install a backend first: pip install tensorflow or pip install torch
CUDA/GPU Issues: Ensure you have the correct GPU drivers and CUDA toolkit installed for your backend.
Memory Issues: For large datasets, consider using JAX backend with GPU support.

Getting Help

If you encounter installation issues:

Check the GitHub Issues
Create a new issue with your system information
Contact the developers: zhangzz@aisi.ac.cn