The registry that provides name -> object mapping, to support third-party users' custom modules. To create a registry (e.g. a backbo

r""" PFT A PyTorch impl of : `Progressive Focused Transformer for Single Image Super-Resolution`. Args: img_size (int | tuple(int

r""" Image to Patch Embedding Args: img_size (int): Image size. Default: 224. patch_size (int): Patch token size. Default: 4.

r""" Image to Patch Unembedding Args: img_size (int): Image size. Default: 224. patch_size (int): Patch token size. Default: 4.

A basic PFT Block for one stage. Args: dim (int): Number of input channels. input_resolution (tuple[int]): Input resolution.

Merge patches into image

Splitting image into patches

Dequantize CbCr channel

CPU prefetcher. Args: loader: Dataloader.

JPEG Quantization for CbCr channels Args: rounding(function): rounding function to use

CUDA prefetcher. Reference: https://github.com/NVIDIA/apex/issues/304# It may consume more GPU memory. Args: loader: Dat

Chroma subsampling on CbCr channels

Upsample chroma layers

Full JPEG compression algorithm Args: rounding(function): rounding function to use

Discrete Cosine Transformation

Full JPEG decompression algorithm Args: rounding(function): rounding function to use

This JPEG algorithm result is slightly different from cv2. DiffJPEG supports batch processing. Args: differentiable(bool): If Tru

Sampler that restricts data loading to a subset of the dataset. Modified from torch.utils.data.distributed.DistributedSampler Support enla

A general file client to access files in different backend. The client loads a file or text in a specified backend from its path and return

Message logger for printing. Args: opt (dict): Config. It contains the following keys: name (str): Exp name.

Adaptive Token Dictionary Block (PFTB). Args: dim (int): Number of input channels. input_resolution (tuple[int]): Input resolutio

r""" PFT Transformer Layer Args: dim (int): Number of input channels. idx (int): Layer index. input_resolution (tuple

Prefetch version of dataloader. Reference: https://github.com/IgorSusmelj/pytorch-styleguide/issues/5# TODO: Need to test on single

A general prefetch generator. Reference: https://stackoverflow.com/questions/7323664/python-generator-pre-fetch Args: generator

Converts RGB image to YCbCr

Upsample module. Args: scale (int): Scale factor. Supported scales: 2^n and 3. num_feat (int): Channel number of intermediate fea

UpsampleOneStep module (the difference with Upsample is that it always only has 1conv + 1pixelshuffle) Used in lightweight SR to save parameter

r""" Shifted Window-based Multi-head Self-Attention (MSA). Args: dim (int): Number of input channels. layer_id (int): Index o

Converts YCbCr image to RGB JPEG

Dequantize Y channel

JPEG Quantization for Y channel Args: rounding(function): rounding function to use

Inverse discrete Cosine Transformation

Base model.

Abstract class of storage backends. All backends need to implement two apis: ``get()`` and ``get_text()``. ``get()`` reads the file as a byt

Charbonnier loss (one variant of Robust L1Loss, a differentiable variant of L1Loss). Described in "Deep Laplacian Pyramid Networks for Fas

Cosine annealing with restarts learning rate scheme. An example of config: periods = [10, 10, 10, 10] restart_weights = [1, 0.5, 0.5

Raw hard disks storage backend.

L1 (mean absolute error, MAE) loss. Args: loss_weight (float): Loss weight for L1 loss. Default: 1.0. reduction (str): Specif

Lmdb storage backend. Args: db_paths (str | list[str]): Lmdb database paths. client_keys (str | list[str]): Lmdb client keys. D

LMDB Maker. Args: lmdb_path (str): Lmdb save path. map_size (int): Map size for lmdb env. Default: 1024 ** 4, 1TB. b

MSE (L2) loss. Args: loss_weight (float): Loss weight for MSE loss. Default: 1.0. reduction (str): Specifies the reduction to

Memcached storage backend. Attributes: server_list_cfg (str): Config file for memcached server list. client_cfg (str): Config f

MultiStep with restarts learning rate scheme. Args: optimizer (torch.nn.optimizer): Torch optimizer. milestones (list): Iter

Paired image dataset for image restoration. Read LQ (Low Quality, e.g. LR (Low Resolution), blurry, noisy, etc) and GT image pairs. Ther

r""" Patch Merging Layer. Args: input_resolution (tuple[int]): Resolution of input feature. dim (int): Number of input channels.

Residual block without BN. Args: num_feat (int): Channel number of intermediate features. Default: 64. res_scale

A custom PyTorch autograd Function for sparse matrix multiplication (SMM) between an activation matrix (A) and a value matrix (V), guided by

A custom PyTorch autograd Function for sparse matrix multiplication (SMM) of query (Q) and key (K) matrices, based on given sparse indices.

Base SR model for single image super-resolution.

Upsample module. Args: scale (int): Scale factor. Supported scales: 2^n and 3. num_feat (int): Channel number of intermediate

Weighted TV loss. Args: loss_weight (float): Loss weight. Default: 1.0.