Symmetric padding same as tensorflow. Ref: https://discuss.pytorch.org/t/symmetric-padding/19866/3

Convert torch Tensors into image numpy arrays. After clamping to [min, max], values will be normalized to [0, 1]. Args: tensor (Tens

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

Apply element-wise weight and reduce loss. Args: loss (Tensor): Element-wise loss. weight (Tensor): Element-wise weights. Default

Args: x: (B, H, W, C) window_size (int): window size Returns: windows: (num_windows*B, window_size, window_size, C)

Args: x: (B, H, W, C) window_size (int): window size Returns: windows: (num_windows*B, window_size, window_size, C)

Args: x: (B, H, W, C) window_size (int): window size Returns: windows: (num_windows*B, window_size, window_size, C)

Args: x: (B, H, W, C) window_size: window size Returns: local window features (num_windows*B, window_size, window_siz

r'''Define parameters for the divisive normalization

Add fog effect to images imgs: numpy array or torch tensor, shape (B, C, H, W) or (C, H, W), range [0, 1]

Degrades a batch of low-light images. img: Tensor of shape (B, C, H, W), where B is the batch size.

Refer to offical code https://github.com/dingkeyan93/DISTS

Args: name (str): the name of this registry

Itialize the paired dataset object for loading and transforming paired data samples from specified dataset folders. This con

default preprocessing of CKDN: https://github.com/researchmm/CKDN Useful when using this metric as losses. Results are slightly differ

It is the training pair pool for increasing the diversity in a batch. Batch processing limits the diversity of synthetic degradations in a ba

It is the training pair pool for increasing the diversity in a batch. Batch processing limits the diversity of synthetic degradations in a ba

Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token first unpad the input, th

Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token first unpad the input, th

Yield items from any nested iterable; see Reference.

Return an activation function given a string

Get the initial lr, which is set by the scheduler.

Get the initial lr, which is set by the scheduler.

Get the initial lr, which is set by the scheduler.

test with non overlap patches

get image path list from image folder

get image path list from lmdb meta info

train with random crop patches

Initialize slurm distributed training environment. If argument ``port`` is not specified, then the master port will be system environment var

Initialize slurm distributed training environment. If argument ``port`` is not specified, then the master port will be system environment var

Initialize slurm distributed training environment. If argument ``port`` is not specified, then the master port will be system environment var

Initialize MUNIT model

Initialize the best metric results dict for recording the best metric value and iteration.