test_pad

deepinv.models.utils.test_pad(model, L, modulo=16, padding_mode='replicate')

Pads the image to fit the model’s expected image size.

Parameters:

• model – model to run on the padded image.

• L (torch.Tensor) – input image.

• modulo (int) – spatial dimensions are padded up to a multiple of modulo.

• padding_mode (str) – padding mode passed to torch.nn.functional.pad(), e.g. "replicate" (default, as in DPIR), "reflect" or "circular". "reflect" and "circular" require the padding to be smaller than the corresponding input dimension; if this is not the case, the padding falls back to "replicate".