PtyRAD performs ptychographic reconstruction using an automatic differentiation (AD) framework powered by PyTorch, enabling flexible and efficient implementation of gradient descent optimization.

PtyRAD integrates various features into the reconstruction, including:

• Mixed-state probe and object, position correction, and position-dependent object tilt correction.
• Allows interoperating with PtychoShelves (fold_slice) and py4DSTEM.
• Provides a set of utility functions to streamline the preprocessing of datasets to facilitate data treatment.

PtyRAD also supports hyperparameter tuning and running on multi-GPU. Additional features include JIT compilation with torch.compile.