For creating seamless tiles.

It will first crop input images to tiles, and then process each tile. Finally, all the processed tiles are merged into one images. Mo

Updates the state with a new batch's gradient statistics, and returns the current gradient noise scale. Args: sq_norm_sma

Updates the state vector with observed bbox.

Zero out the parameters of a module and return it.

Zero out the parameters of a module and return it.

Args: med_config (str): path for the mixture of encoder-decoder model's configuration file image_size (int): input im

Assert that the shapes of the sub-modules in `source_order` and `target_order` are aligned.

Convert the source model's state_dict to match the target model's state_dict.

Generates decreasing timesteps with 'leading' spacing and offset of 1 similar to diffusers settings for the DDIM scheduler in Stable

Generates a tensor of timesteps. This method should be overridden by subclasses to provide the specific timesteps for the diffusion

Return an activation function given a string.

Load a diffusion model. Weights location may also be shortcut name, e.g. "SD-1.5"

Check if a module is a leaf module with weights.

Load weights from .npz checkpoints for official Google Brain Flax implementation.