sometimes shit gets messed up. We try to catch that here

output_file must end with .json; can be None

used for debugging plans etc

if output_dtype is None then we use np.uint8/torch.uint8 if input is torch.Tensor then output will be on the same device np.ndarray is f

Generates a dataset.json file in the output folder channel_names: Channel names must map the index to the name of the channel, examp

This is nnU-Net's default function for making predictions. It works best for batch predictions (predicting many images at once).

This function is used to set the number of processes used on different Systems. It is specific to our cluster infrastructure at DKFZ. You can

This is used when making predictions with a trained model

returns True if the number of results that are not ready is greater than the number of available workers + allowed_num_queued

This function is stupid and certainly one of the weakest spots of this implementation. Not entirely sure how we can fix it.

Compute the output size given input size and target long side length.

Merge overlay key-value pairs on top of base preprocess cfg or dict. Input dicts are filtered based on PreprocessCfg fields.

each element returned by data_iterator must be a dict with 'data', 'ofile' and 'data_properites' keys! If 'ofile' is None, the result

Predict masks for the given input prompts, using the currently set image. Input prompts are batched torch tensors and are expected

Replace a given module within a parent module with some third module Useful for injecting new layers in an existing model.

seg file can be none (test cases) order of operations is: transpose -> crop -> resample so when we export we need to run the

Args: dim (int): Number of input channels. num_heads (int): Number of attention heads in each ViT block.