import gdown # url = 'https://drive.google.com/file/d/1LHIUM7YoUDk8cXWzVZhroAcA1xXi-d87/view?usp=drive_link' output = 'models/HAT/hat_model_checkpoint_best.pth' # gdown.download(url, output, quiet=False) import gc import os import random import time import wandb from tqdm import tqdm import matplotlib.pyplot as plt from PIL import Image from skimage.metrics import structural_similarity as ssim import torch from torch import nn, optim import torch.nn.functional as F from torch.utils.data import Dataset, DataLoader, ConcatDataset from torchvision import transforms from torchvision.transforms import Compose from torchmetrics.functional.image import structural_similarity_index_measure as ssim from basicsr.archs.arch_util import to_2tuple, trunc_normal_ from einops import rearrange import math class ChannelAttention(nn.Module): """Channel attention used in RCAN. Args: num_feat (int): Channel number of intermediate features. squeeze_factor (int): Channel squeeze factor. Default: 16. """ def __init__(self, num_feat, squeeze_factor=16): super(ChannelAttention, self).__init__() self.attention = nn.Sequential( nn.AdaptiveAvgPool2d(1), nn.Conv2d(num_feat, num_feat // squeeze_factor, 1, padding=0), nn.ReLU(inplace=True), nn.Conv2d(num_feat // squeeze_factor, num_feat, 1, padding=0), nn.Sigmoid()) def forward(self, x): y = self.attention(x) return x * y class CAB(nn.Module): def __init__(self, num_feat, compress_ratio=3, squeeze_factor=30): super(CAB, self).__init__() self.cab = nn.Sequential( nn.Conv2d(num_feat, num_feat // compress_ratio, 3, 1, 1), nn.GELU(), nn.Conv2d(num_feat // compress_ratio, num_feat, 3, 1, 1), ChannelAttention(num_feat, squeeze_factor) ) def forward(self, x): return self.cab(x) def window_partition(x, window_size): """ Args: x: (b, h, w, c) window_size (int): window size Returns: windows: (num_windows*b, window_size, window_size, c) """ b, h, w, c = x.shape x = x.view(b, h // window_size, window_size, w // window_size, window_size, c) windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, c) return windows def window_reverse(windows, window_size, h, w): """ Args: windows: (num_windows*b, window_size, window_size, c) window_size (int): Window size h (int): Height of image w (int): Width of image Returns: x: (b, h, w, c) """ b = int(windows.shape[0] / (h * w / window_size / window_size)) x = windows.view(b, h // window_size, w // window_size, window_size, window_size, -1) x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(b, h, w, -1) return x class WindowAttention(nn.Module): r""" Window based multi-head self attention (W-MSA) module with relative position bias. It supports both of shifted and non-shifted window. Args: dim (int): Number of input channels. window_size (tuple[int]): The height and width of the window. num_heads (int): Number of attention heads. qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0 proj_drop (float, optional): Dropout ratio of output. Default: 0.0 """ def __init__(self, dim, window_size, num_heads, qkv_bias=True, qk_scale=None, attn_drop=0., proj_drop=0.): super().__init__() self.dim = dim self.window_size = window_size # Wh, Ww self.num_heads = num_heads head_dim = dim // num_heads self.scale = qk_scale or head_dim**-0.5 # define a parameter table of relative position bias self.relative_position_bias_table = nn.Parameter( torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads)) # 2*Wh-1 * 2*Ww-1, nH self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) self.attn_drop = nn.Dropout(attn_drop) self.proj = nn.Linear(dim, dim) self.proj_drop = nn.Dropout(proj_drop) trunc_normal_(self.relative_position_bias_table, std=.02) self.softmax = nn.Softmax(dim=-1) def forward(self, x, rpi, mask=None): """ Args: x: input features with shape of (num_windows*b, n, c) mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None """ b_, n, c = x.shape qkv = self.qkv(x).reshape(b_, n, 3, self.num_heads, c // self.num_heads).permute(2, 0, 3, 1, 4) q, k, v = qkv[0], qkv[1], qkv[2] # make torchscript happy (cannot use tensor as tuple) q = q * self.scale attn = (q @ k.transpose(-2, -1)) relative_position_bias = self.relative_position_bias_table[rpi.view(-1)].view( self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1) # Wh*Ww,Wh*Ww,nH relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous() # nH, Wh*Ww, Wh*Ww attn = attn + relative_position_bias.unsqueeze(0) if mask is not None: nw = mask.shape[0] attn = attn.view(b_ // nw, nw, self.num_heads, n, n) + mask.unsqueeze(1).unsqueeze(0) attn = attn.view(-1, self.num_heads, n, n) attn = self.softmax(attn) else: attn = self.softmax(attn) attn = self.attn_drop(attn) x = (attn @ v).transpose(1, 2).reshape(b_, n, c) x = self.proj(x) x = self.proj_drop(x) return x def drop_path(x, drop_prob: float = 0., training: bool = False): """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). From: https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/layers/drop.py """ if drop_prob == 0. or not training: return x keep_prob = 1 - drop_prob shape = (x.shape[0], ) + (1, ) * (x.ndim - 1) # work with diff dim tensors, not just 2D ConvNets random_tensor = keep_prob + torch.rand(shape, dtype=x.dtype, device=x.device) random_tensor.floor_() # binarize output = x.div(keep_prob) * random_tensor return output class DropPath(nn.Module): """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). From: https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/layers/drop.py """ def __init__(self, drop_prob=None): super(DropPath, self).__init__() self.drop_prob = drop_prob def forward(self, x): return drop_path(x, self.drop_prob, self.training) class Mlp(nn.Module): def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.): super().__init__() out_features = out_features or in_features hidden_features = hidden_features or in_features self.fc1 = nn.Linear(in_features, hidden_features) self.act = act_layer() self.fc2 = nn.Linear(hidden_features, out_features) self.drop = nn.Dropout(drop) def forward(self, x): x = self.fc1(x) x = self.act(x) x = self.drop(x) x = self.fc2(x) x = self.drop(x) return x class OCAB(nn.Module): # overlapping cross-attention block def __init__(self, dim, input_resolution, window_size, overlap_ratio, num_heads, qkv_bias=True, qk_scale=None, mlp_ratio=2, norm_layer=nn.LayerNorm ): super().__init__() self.dim = dim self.input_resolution = input_resolution self.window_size = window_size self.num_heads = num_heads head_dim = dim // num_heads self.scale = qk_scale or head_dim**-0.5 self.overlap_win_size = int(window_size * overlap_ratio) + window_size self.norm1 = norm_layer(dim) self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) self.unfold = nn.Unfold(kernel_size=(self.overlap_win_size, self.overlap_win_size), stride=window_size, padding=(self.overlap_win_size-window_size)//2) # define a parameter table of relative position bias self.relative_position_bias_table = nn.Parameter( torch.zeros((window_size + self.overlap_win_size - 1) * (window_size + self.overlap_win_size - 1), num_heads)) # 2*Wh-1 * 2*Ww-1, nH trunc_normal_(self.relative_position_bias_table, std=.02) self.softmax = nn.Softmax(dim=-1) self.proj = nn.Linear(dim,dim) self.norm2 = norm_layer(dim) mlp_hidden_dim = int(dim * mlp_ratio) self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=nn.GELU) def forward(self, x, x_size, rpi): h, w = x_size b, _, c = x.shape shortcut = x x = self.norm1(x) x = x.view(b, h, w, c) qkv = self.qkv(x).reshape(b, h, w, 3, c).permute(3, 0, 4, 1, 2) # 3, b, c, h, w q = qkv[0].permute(0, 2, 3, 1) # b, h, w, c kv = torch.cat((qkv[1], qkv[2]), dim=1) # b, 2*c, h, w # partition windows q_windows = window_partition(q, self.window_size) # nw*b, window_size, window_size, c q_windows = q_windows.view(-1, self.window_size * self.window_size, c) # nw*b, window_size*window_size, c kv_windows = self.unfold(kv) # b, c*w*w, nw kv_windows = rearrange(kv_windows, 'b (nc ch owh oww) nw -> nc (b nw) (owh oww) ch', nc=2, ch=c, owh=self.overlap_win_size, oww=self.overlap_win_size).contiguous() # 2, nw*b, ow*ow, c k_windows, v_windows = kv_windows[0], kv_windows[1] # nw*b, ow*ow, c b_, nq, _ = q_windows.shape _, n, _ = k_windows.shape d = self.dim // self.num_heads q = q_windows.reshape(b_, nq, self.num_heads, d).permute(0, 2, 1, 3) # nw*b, nH, nq, d k = k_windows.reshape(b_, n, self.num_heads, d).permute(0, 2, 1, 3) # nw*b, nH, n, d v = v_windows.reshape(b_, n, self.num_heads, d).permute(0, 2, 1, 3) # nw*b, nH, n, d q = q * self.scale attn = (q @ k.transpose(-2, -1)) relative_position_bias = self.relative_position_bias_table[rpi.view(-1)].view( self.window_size * self.window_size, self.overlap_win_size * self.overlap_win_size, -1) # ws*ws, wse*wse, nH relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous() # nH, ws*ws, wse*wse attn = attn + relative_position_bias.unsqueeze(0) attn = self.softmax(attn) attn_windows = (attn @ v).transpose(1, 2).reshape(b_, nq, self.dim) # merge windows attn_windows = attn_windows.view(-1, self.window_size, self.window_size, self.dim) x = window_reverse(attn_windows, self.window_size, h, w) # b h w c x = x.view(b, h * w, self.dim) x = self.proj(x) + shortcut x = x + self.mlp(self.norm2(x)) return x class AttenBlocks(nn.Module): """ A series of attention blocks for one RHAG. Args: dim (int): Number of input channels. input_resolution (tuple[int]): Input resolution. depth (int): Number of blocks. num_heads (int): Number of attention heads. window_size (int): Local window size. mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set. drop (float, optional): Dropout rate. Default: 0.0 attn_drop (float, optional): Attention dropout rate. Default: 0.0 drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0 norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. """ def __init__(self, dim, input_resolution, depth, num_heads, window_size, compress_ratio, squeeze_factor, conv_scale, overlap_ratio, mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False): super().__init__() self.dim = dim self.input_resolution = input_resolution self.depth = depth self.use_checkpoint = use_checkpoint # build blocks self.blocks = nn.ModuleList([ HAB( dim=dim, input_resolution=input_resolution, num_heads=num_heads, window_size=window_size, shift_size=0 if (i % 2 == 0) else window_size // 2, compress_ratio=compress_ratio, squeeze_factor=squeeze_factor, conv_scale=conv_scale, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop, attn_drop=attn_drop, drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path, norm_layer=norm_layer) for i in range(depth) ]) # OCAB self.overlap_attn = OCAB( dim=dim, input_resolution=input_resolution, window_size=window_size, overlap_ratio=overlap_ratio, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, mlp_ratio=mlp_ratio, norm_layer=norm_layer ) # patch merging layer if downsample is not None: self.downsample = downsample(input_resolution, dim=dim, norm_layer=norm_layer) else: self.downsample = None def forward(self, x, x_size, params): for blk in self.blocks: x = blk(x, x_size, params['rpi_sa'], params['attn_mask']) x = self.overlap_attn(x, x_size, params['rpi_oca']) if self.downsample is not None: x = self.downsample(x) return x class RHAG(nn.Module): """Residual Hybrid Attention Group (RHAG). Args: dim (int): Number of input channels. input_resolution (tuple[int]): Input resolution. depth (int): Number of blocks. num_heads (int): Number of attention heads. window_size (int): Local window size. mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set. drop (float, optional): Dropout rate. Default: 0.0 attn_drop (float, optional): Attention dropout rate. Default: 0.0 drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0 norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. img_size: Input image size. patch_size: Patch size. resi_connection: The convolutional block before residual connection. """ def __init__(self, dim, input_resolution, depth, num_heads, window_size, compress_ratio, squeeze_factor, conv_scale, overlap_ratio, mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False, img_size=224, patch_size=4, resi_connection='1conv'): super(RHAG, self).__init__() self.dim = dim self.input_resolution = input_resolution self.residual_group = AttenBlocks( dim=dim, input_resolution=input_resolution, depth=depth, num_heads=num_heads, window_size=window_size, compress_ratio=compress_ratio, squeeze_factor=squeeze_factor, conv_scale=conv_scale, overlap_ratio=overlap_ratio, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop, attn_drop=attn_drop, drop_path=drop_path, norm_layer=norm_layer, downsample=downsample, use_checkpoint=use_checkpoint) if resi_connection == '1conv': self.conv = nn.Conv2d(dim, dim, 3, 1, 1) elif resi_connection == 'identity': self.conv = nn.Identity() self.patch_embed = PatchEmbed( img_size=img_size, patch_size=patch_size, in_chans=0, embed_dim=dim, norm_layer=None) self.patch_unembed = PatchUnEmbed( img_size=img_size, patch_size=patch_size, in_chans=0, embed_dim=dim, norm_layer=None) def forward(self, x, x_size, params): return self.patch_embed(self.conv(self.patch_unembed(self.residual_group(x, x_size, params), x_size))) + x class PatchEmbed(nn.Module): r""" Image to Patch Embedding Args: img_size (int): Image size. Default: 224. patch_size (int): Patch token size. Default: 4. in_chans (int): Number of input image channels. Default: 3. embed_dim (int): Number of linear projection output channels. Default: 96. norm_layer (nn.Module, optional): Normalization layer. Default: None """ def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None): super().__init__() img_size = to_2tuple(img_size) patch_size = to_2tuple(patch_size) patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]] self.img_size = img_size self.patch_size = patch_size self.patches_resolution = patches_resolution self.num_patches = patches_resolution[0] * patches_resolution[1] self.in_chans = in_chans self.embed_dim = embed_dim if norm_layer is not None: self.norm = norm_layer(embed_dim) else: self.norm = None def forward(self, x): x = x.flatten(2).transpose(1, 2) # b Ph*Pw c if self.norm is not None: x = self.norm(x) return x class PatchUnEmbed(nn.Module): r""" Image to Patch Unembedding Args: img_size (int): Image size. Default: 224. patch_size (int): Patch token size. Default: 4. in_chans (int): Number of input image channels. Default: 3. embed_dim (int): Number of linear projection output channels. Default: 96. norm_layer (nn.Module, optional): Normalization layer. Default: None """ def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None): super().__init__() img_size = to_2tuple(img_size) patch_size = to_2tuple(patch_size) patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]] self.img_size = img_size self.patch_size = patch_size self.patches_resolution = patches_resolution self.num_patches = patches_resolution[0] * patches_resolution[1] self.in_chans = in_chans self.embed_dim = embed_dim def forward(self, x, x_size): x = x.transpose(1, 2).contiguous().view(x.shape[0], self.embed_dim, x_size[0], x_size[1]) # b Ph*Pw c return x class Upsample(nn.Sequential): """Upsample module. Args: scale (int): Scale factor. Supported scales: 2^n and 3. num_feat (int): Channel number of intermediate features. """ def __init__(self, scale, num_feat): m = [] if (scale & (scale - 1)) == 0: # scale = 2^n for _ in range(int(math.log(scale, 2))): m.append(nn.Conv2d(num_feat, 4 * num_feat, 3, 1, 1)) m.append(nn.PixelShuffle(2)) elif scale == 3: m.append(nn.Conv2d(num_feat, 9 * num_feat, 3, 1, 1)) m.append(nn.PixelShuffle(3)) else: raise ValueError(f'scale {scale} is not supported. ' 'Supported scales: 2^n and 3.') super(Upsample, self).__init__(*m) class HAT(nn.Module): r""" Hybrid Attention Transformer A PyTorch implementation of : `Activating More Pixels in Image Super-Resolution Transformer`. Some codes are based on SwinIR. Args: img_size (int | tuple(int)): Input image size. Default 64 patch_size (int | tuple(int)): Patch size. Default: 1 in_chans (int): Number of input image channels. Default: 3 embed_dim (int): Patch embedding dimension. Default: 96 depths (tuple(int)): Depth of each Swin Transformer layer. num_heads (tuple(int)): Number of attention heads in different layers. window_size (int): Window size. Default: 7 mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4 qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True qk_scale (float): Override default qk scale of head_dim ** -0.5 if set. Default: None drop_rate (float): Dropout rate. Default: 0 attn_drop_rate (float): Attention dropout rate. Default: 0 drop_path_rate (float): Stochastic depth rate. Default: 0.1 norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm. ape (bool): If True, add absolute position embedding to the patch embedding. Default: False patch_norm (bool): If True, add normalization after patch embedding. Default: True use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False upscale: Upscale factor. 2/3/4/8 for image SR, 1 for denoising and compress artifact reduction img_range: Image range. 1. or 255. upsampler: The reconstruction reconstruction module. 'pixelshuffle'/'pixelshuffledirect'/'nearest+conv'/None resi_connection: The convolutional block before residual connection. '1conv'/'3conv' """ def __init__(self, img_size=64, patch_size=1, in_chans=3, embed_dim=96, depths=(6, 6, 6, 6), num_heads=(6, 6, 6, 6), window_size=7, compress_ratio=3, squeeze_factor=30, conv_scale=0.01, overlap_ratio=0.5, mlp_ratio=4., qkv_bias=True, qk_scale=None, drop_rate=0., attn_drop_rate=0., drop_path_rate=0.1, norm_layer=nn.LayerNorm, ape=False, patch_norm=True, use_checkpoint=False, upscale=2, img_range=1., upsampler='', resi_connection='1conv', **kwargs): super(HAT, self).__init__() self.window_size = window_size self.shift_size = window_size // 2 self.overlap_ratio = overlap_ratio num_in_ch = in_chans num_out_ch = in_chans num_feat = 64 self.img_range = img_range if in_chans == 3: rgb_mean = (0.4488, 0.4371, 0.4040) self.mean = torch.Tensor(rgb_mean).view(1, 3, 1, 1) else: self.mean = torch.zeros(1, 1, 1, 1) self.upscale = upscale self.upsampler = upsampler # relative position index relative_position_index_SA = self.calculate_rpi_sa() relative_position_index_OCA = self.calculate_rpi_oca() self.register_buffer('relative_position_index_SA', relative_position_index_SA) self.register_buffer('relative_position_index_OCA', relative_position_index_OCA) # ------------------------- 1, shallow feature extraction ------------------------- # self.conv_first = nn.Conv2d(num_in_ch, embed_dim, 3, 1, 1) # ------------------------- 2, deep feature extraction ------------------------- # self.num_layers = len(depths) self.embed_dim = embed_dim self.ape = ape self.patch_norm = patch_norm self.num_features = embed_dim self.mlp_ratio = mlp_ratio # split image into non-overlapping patches self.patch_embed = PatchEmbed( img_size=img_size, patch_size=patch_size, in_chans=embed_dim, embed_dim=embed_dim, norm_layer=norm_layer if self.patch_norm else None) num_patches = self.patch_embed.num_patches patches_resolution = self.patch_embed.patches_resolution self.patches_resolution = patches_resolution # merge non-overlapping patches into image self.patch_unembed = PatchUnEmbed( img_size=img_size, patch_size=patch_size, in_chans=embed_dim, embed_dim=embed_dim, norm_layer=norm_layer if self.patch_norm else None) # absolute position embedding if self.ape: self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim)) trunc_normal_(self.absolute_pos_embed, std=.02) self.pos_drop = nn.Dropout(p=drop_rate) # stochastic depth dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))] # stochastic depth decay rule # build Residual Hybrid Attention Groups (RHAG) self.layers = nn.ModuleList() for i_layer in range(self.num_layers): layer = RHAG( dim=embed_dim, input_resolution=(patches_resolution[0], patches_resolution[1]), depth=depths[i_layer], num_heads=num_heads[i_layer], window_size=window_size, compress_ratio=compress_ratio, squeeze_factor=squeeze_factor, conv_scale=conv_scale, overlap_ratio=overlap_ratio, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])], # no impact on SR results norm_layer=norm_layer, downsample=None, use_checkpoint=use_checkpoint, img_size=img_size, patch_size=patch_size, resi_connection=resi_connection) self.layers.append(layer) self.norm = norm_layer(self.num_features) # build the last conv layer in deep feature extraction if resi_connection == '1conv': self.conv_after_body = nn.Conv2d(embed_dim, embed_dim, 3, 1, 1) elif resi_connection == 'identity': self.conv_after_body = nn.Identity() # ------------------------- 3, high quality image reconstruction ------------------------- # if self.upsampler == 'pixelshuffle': # for classical SR self.conv_before_upsample = nn.Sequential( nn.Conv2d(embed_dim, num_feat, 3, 1, 1), nn.LeakyReLU(inplace=True)) self.upsample = Upsample(upscale, num_feat) self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) self.apply(self._init_weights) def _init_weights(self, m): if isinstance(m, nn.Linear): trunc_normal_(m.weight, std=.02) if isinstance(m, nn.Linear) and m.bias is not None: nn.init.constant_(m.bias, 0) elif isinstance(m, nn.LayerNorm): nn.init.constant_(m.bias, 0) nn.init.constant_(m.weight, 1.0) def calculate_rpi_sa(self): # calculate relative position index for SA coords_h = torch.arange(self.window_size) coords_w = torch.arange(self.window_size) coords = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, Wh, Ww coords_flatten = torch.flatten(coords, 1) # 2, Wh*Ww relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :] # 2, Wh*Ww, Wh*Ww relative_coords = relative_coords.permute(1, 2, 0).contiguous() # Wh*Ww, Wh*Ww, 2 relative_coords[:, :, 0] += self.window_size - 1 # shift to start from 0 relative_coords[:, :, 1] += self.window_size - 1 relative_coords[:, :, 0] *= 2 * self.window_size - 1 relative_position_index = relative_coords.sum(-1) # Wh*Ww, Wh*Ww return relative_position_index def calculate_rpi_oca(self): # calculate relative position index for OCA window_size_ori = self.window_size window_size_ext = self.window_size + int(self.overlap_ratio * self.window_size) coords_h = torch.arange(window_size_ori) coords_w = torch.arange(window_size_ori) coords_ori = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, ws, ws coords_ori_flatten = torch.flatten(coords_ori, 1) # 2, ws*ws coords_h = torch.arange(window_size_ext) coords_w = torch.arange(window_size_ext) coords_ext = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, wse, wse coords_ext_flatten = torch.flatten(coords_ext, 1) # 2, wse*wse relative_coords = coords_ext_flatten[:, None, :] - coords_ori_flatten[:, :, None] # 2, ws*ws, wse*wse relative_coords = relative_coords.permute(1, 2, 0).contiguous() # ws*ws, wse*wse, 2 relative_coords[:, :, 0] += window_size_ori - window_size_ext + 1 # shift to start from 0 relative_coords[:, :, 1] += window_size_ori - window_size_ext + 1 relative_coords[:, :, 0] *= window_size_ori + window_size_ext - 1 relative_position_index = relative_coords.sum(-1) return relative_position_index def calculate_mask(self, x_size): # calculate attention mask for SW-MSA h, w = x_size img_mask = torch.zeros((1, h, w, 1)) # 1 h w 1 h_slices = (slice(0, -self.window_size), slice(-self.window_size, -self.shift_size), slice(-self.shift_size, None)) w_slices = (slice(0, -self.window_size), slice(-self.window_size, -self.shift_size), slice(-self.shift_size, None)) cnt = 0 for h in h_slices: for w in w_slices: img_mask[:, h, w, :] = cnt cnt += 1 mask_windows = window_partition(img_mask, self.window_size) # nw, window_size, window_size, 1 mask_windows = mask_windows.view(-1, self.window_size * self.window_size) attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2) attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0)) return attn_mask @torch.jit.ignore def no_weight_decay(self): return {'absolute_pos_embed'} @torch.jit.ignore def no_weight_decay_keywords(self): return {'relative_position_bias_table'} def forward_features(self, x): x_size = (x.shape[2], x.shape[3]) # Calculate attention mask and relative position index in advance to speed up inference. # The original code is very time-consuming for large window size. attn_mask = self.calculate_mask(x_size).to(x.device) params = {'attn_mask': attn_mask, 'rpi_sa': self.relative_position_index_SA, 'rpi_oca': self.relative_position_index_OCA} x = self.patch_embed(x) if self.ape: x = x + self.absolute_pos_embed x = self.pos_drop(x) for layer in self.layers: x = layer(x, x_size, params) x = self.norm(x) # b seq_len c x = self.patch_unembed(x, x_size) return x def forward(self, x): self.mean = self.mean.type_as(x) x = (x - self.mean) * self.img_range if self.upsampler == 'pixelshuffle': # for classical SR x = self.conv_first(x) x = self.conv_after_body(self.forward_features(x)) + x x = self.conv_before_upsample(x) x = self.conv_last(self.upsample(x)) x = x / self.img_range + self.mean return x # ------------------------------ HYPERPARAMS ------------------------------ # config = { "network_g": { "type": "HAT", "upscale": 4, "in_chans": 3, "img_size": 64, "window_size": 16, "compress_ratio": 3, "squeeze_factor": 30, "conv_scale": 0.01, "overlap_ratio": 0.5, "img_range": 1., "depths": [6, 6, 6, 6, 6, 6], "embed_dim": 180, "num_heads": [6, 6, 6, 6, 6, 6], "mlp_ratio": 2, "upsampler": 'pixelshuffle', "resi_connection": '1conv' }, "train": { "ema_decay": 0.999, "optim_g": { "type": "Adam", "lr": 1e-4, "weight_decay": 0, "betas": [0.9, 0.99] }, "scheduler": { "type": "MultiStepLR", "milestones": [12, 20, 25, 30], "gamma": 0.5 }, "total_iter": 30, "warmup_iter": -1, "pixel_opt": { "type": "L1Loss", "loss_weight": 1.0, "reduction": "mean" } }, 'tile':{ 'tile_size': 56, 'tile_pad': 4 } } DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') DEVICE class Network: def __init__(self, train_dataloader=train_dataloader, valid_dataloader=valid_dataloader, config = config, device=DEVICE, run_id=None, wandb_mode = False, STOP = float('inf'), save_temp_model = True, train_model_continue = False): self.config = config self.model = HAT( upscale=self.config['network_g']['upscale'], in_chans=self.config['network_g']['in_chans'], img_size=self.config['network_g']['img_size'], window_size=self.config['network_g']['window_size'], compress_ratio=self.config['network_g']['compress_ratio'], squeeze_factor=self.config['network_g']['squeeze_factor'], conv_scale=self.config['network_g']['conv_scale'], overlap_ratio=self.config['network_g']['overlap_ratio'], img_range=self.config['network_g']['img_range'], depths=self.config['network_g']['depths'], embed_dim=self.config['network_g']['embed_dim'], num_heads=self.config['network_g']['num_heads'], mlp_ratio=self.config['network_g']['mlp_ratio'], upsampler=self.config['network_g']['upsampler'], resi_connection=self.config['network_g']['resi_connection'] ).to(device) self.device = device self.STOP = STOP self.wandb_mode = wandb_mode self.loss_fn = nn.L1Loss(reduction='mean').to(device) self.optimizer = optim.Adam(self.model.parameters(), lr=self.config['train']['optim_g']['lr'], weight_decay=config['train']['optim_g']['weight_decay'],betas=tuple(config['train']['optim_g']['betas'])) self.scheduler = optim.lr_scheduler.MultiStepLR(self.optimizer, milestones = self.config['train']['scheduler']['milestones'], gamma=self.config['train']['scheduler']['gamma']) self.train_dataloader = train_dataloader self.valid_dataloader = valid_dataloader self.num_epochs = self.config['train']['total_iter'] self.run_id = run_id self.save_temp_model = save_temp_model self.train_model_continue = train_model_continue self.last_valid_loss = float('inf') checkpoint_path = output if self.save_temp_model: if self.train_model_continue: # Load the network and other states from the checkpoint self.start_epoch, train_loss, valid_loss = self.load_network(checkpoint_path) initial_lr = self.config['train']['optim_g']['lr'] * self.config['train']['scheduler']['gamma'] # Define your initial or desired learning rate for param_group in self.optimizer.param_groups: param_group['lr'] = initial_lr # Resetting learning rate # Recreate the scheduler with the updated optimizer self.scheduler = optim.lr_scheduler.MultiStepLR( self.optimizer, milestones=self.config['train']['scheduler']['milestones'], gamma=self.config['train']['scheduler']['gamma'], last_epoch = self.start_epoch - 1 # Ensure to set the last_epoch to continue correctly ) # Print the updated learning rate and scheduler state print("Updated Learning Rate is:", self.optimizer.param_groups[0]['lr']) print(self.scheduler.state_dict()) self.last_valid_loss = valid_loss # self.num_epochs-= self.start_epoch print("Previous train loss: ", train_loss) print("Previous valid loss: ", self.last_valid_loss) # Resume training notice print("------------------- Resuming training -------------------") self.save_network(0, 0, 0, 'temp_model_checkpoint.pth') def del_model(self): del self.model del self.optimizer del self.scheduler gc.collect() torch.cuda.empty_cache() def pre_process(self): # pad to multiplication of window_size window_size = self.config['network_g']['window_size'] * 4 self.scale = self.config['network_g']['upscale'] self.mod_pad_h, self.mod_pad_w = 0, 0 _, _, h, w = self.input_tile.size() if h % window_size != 0: self.mod_pad_h = window_size - h % window_size # Loop to add padding to the height until it's a multiple of window_size for i in range(self.mod_pad_h): self.input_tile = F.pad(self.input_tile, (0, 0, 0, 1), 'reflect') if w % window_size != 0: # Loop to add padding to the width until it's a multiple of window_size self.mod_pad_w = window_size - w % window_size for i in range(self.mod_pad_w): self.input_tile = F.pad(self.input_tile, (0, 1, 0, 0), 'reflect') def post_process(self): _, _, h, w = self.output_tile.size() self.output_tile = self.output_tile[:, :, 0:h - self.mod_pad_h * self.scale, 0:w - self.mod_pad_w * self.scale] def save_network(self, epoch, train_loss, valid_loss, checkpoint_path): checkpoint = { 'epoch': epoch, 'train_loss': train_loss, 'valid_loss': valid_loss, 'model': self.model.state_dict(), 'optimizer': self.optimizer.state_dict(), 'learning_rate_scheduler': self.scheduler.state_dict(), 'network': self } torch.save(checkpoint, checkpoint_path) def load_network(self, checkpoint_path): checkpoint = torch.load(checkpoint_path, map_location=self.device) self.model = HAT( upscale=self.config['network_g']['upscale'], in_chans=self.config['network_g']['in_chans'], img_size=self.config['network_g']['img_size'], window_size=self.config['network_g']['window_size'], compress_ratio=self.config['network_g']['compress_ratio'], squeeze_factor=self.config['network_g']['squeeze_factor'], conv_scale=self.config['network_g']['conv_scale'], overlap_ratio=self.config['network_g']['overlap_ratio'], img_range=self.config['network_g']['img_range'], depths=self.config['network_g']['depths'], embed_dim=self.config['network_g']['embed_dim'], num_heads=self.config['network_g']['num_heads'], mlp_ratio=self.config['network_g']['mlp_ratio'], upsampler=self.config['network_g']['upsampler'], resi_connection=self.config['network_g']['resi_connection'] ).to(self.device) self.optimizer = optim.Adam(self.model.parameters(), lr=self.config['train']['optim_g']['lr'], weight_decay=config['train']['optim_g']['weight_decay'],betas=tuple(config['train']['optim_g']['betas'])) self.model.load_state_dict(checkpoint['model']) self.optimizer.load_state_dict(checkpoint['optimizer']) # before create and load scheduler self.scheduler = optim.lr_scheduler.MultiStepLR(self.optimizer, milestones = self.config['train']['scheduler']['milestones'], gamma=self.config['train']['scheduler']['gamma']) self.scheduler.load_state_dict(checkpoint['learning_rate_scheduler']) return checkpoint['epoch'], checkpoint['train_loss'], checkpoint['valid_loss'] def train_step(self, lr_images, hr_images): lr_images, hr_images = lr_images.to(self.device), hr_images.to(self.device) sr_images = self.model(lr_images) self.optimizer.zero_grad() loss = self.loss_fn(sr_images, hr_images) loss.backward() self.optimizer.step() # Memory cleanup del sr_images, lr_images, hr_images gc.collect() torch.cuda.empty_cache() return loss.item() def valid_step(self, lr_images, hr_images): lr_images, hr_images = lr_images.to(self.device), hr_images.to(self.device) sr_images = self.tile_valid(lr_images) loss = self.loss_fn(sr_images, hr_images) # Memory cleanup del sr_images, lr_images, hr_images gc.collect() torch.cuda.empty_cache() return loss.item() def tile_valid(self, lr_images): """ Process all tiles of an image in a batch and then merge them back into the output image. """ batch, channel, height, width = lr_images.shape output_height = height * self.config['network_g']['upscale'] output_width = width * self.config['network_g']['upscale'] output_shape = (batch, channel, output_height, output_width) # Start with black image for output sr_images = lr_images.new_zeros(output_shape) tiles_x = math.ceil(width / self.config['tile']['tile_size']) tiles_y = math.ceil(height / self.config['tile']['tile_size']) tile_list = [] # Extract all tiles for y in range(tiles_y): for x in range(tiles_x): input_start_x = x * self.config['tile']['tile_size'] input_end_x = min(input_start_x + self.config['tile']['tile_size'], width) input_start_y = y * self.config['tile']['tile_size'] input_end_y = min(input_start_y + self.config['tile']['tile_size'], height) input_start_x_pad = max(input_start_x - self.config['tile']['tile_pad'], 0) input_end_x_pad = min(input_end_x + self.config['tile']['tile_pad'], width) input_start_y_pad = max(input_start_y - self.config['tile']['tile_pad'], 0) input_end_y_pad = min(input_end_y + self.config['tile']['tile_pad'], height) # Extract tile and add to list self.input_tile = lr_images[:, :, input_start_y_pad:input_end_y_pad, input_start_x_pad:input_end_x_pad] self.pre_process() tile_list.append(self.input_tile.clone()) output_tiles = [] # Determine the number of tiles to process per batch batch_size = 16 # Adjust based on your specific situation for i in range(0, len(tile_list), batch_size): # Extract a batch of tiles batch = tile_list[i:i + batch_size] tile_batch = torch.cat(batch, dim=0) # This creates a batch of tiles # Process the batch through the model self.model.eval() with torch.no_grad(): # Ensure that each tile processed by the model returns a 3D tensor (C, H, W) output_batch = self.model(tile_batch) # Extend the list of processed tiles output_tiles.append(output_batch) # Assuming output_batch is 4D # Concatenate along the first dimension to combine all the processed tiles output_tile_batch = torch.cat(output_tiles, dim=0) # This should be 4D now for y in range(tiles_y): for x in range(tiles_x): # input tile area on total image input_start_x = x * self.config['tile']['tile_size'] input_end_x = min(input_start_x + self.config['tile']['tile_size'], width) input_start_y = y * self.config['tile']['tile_size'] input_end_y = min(input_start_y + self.config['tile']['tile_size'], height) # input tile area on total image with padding input_start_x_pad = max(input_start_x - self.config['tile']['tile_pad'], 0) input_end_x_pad = min(input_end_x + self.config['tile']['tile_pad'], width) input_start_y_pad = max(input_start_y - self.config['tile']['tile_pad'], 0) input_end_y_pad = min(input_end_y + self.config['tile']['tile_pad'], height) # input tile dimensions input_tile_width = input_end_x - input_start_x input_tile_height = input_end_y - input_start_y tile_idx = y * tiles_x + x self.pre_process() self.output_tile = output_tile_batch[tile_idx, :, :, :].unsqueeze(0).clone() self.post_process() # output tile area on total image output_start_x = input_start_x * self.config['network_g']['upscale'] output_end_x = input_end_x * self.config['network_g']['upscale'] output_start_y = input_start_y * self.config['network_g']['upscale'] output_end_y = input_end_y * self.config['network_g']['upscale'] # output tile area without padding output_start_x_tile = (input_start_x - input_start_x_pad) * self.config['network_g']['upscale'] output_end_x_tile = output_start_x_tile + input_tile_width * self.config['network_g']['upscale'] output_start_y_tile = (input_start_y - input_start_y_pad) * self.config['network_g']['upscale'] output_end_y_tile = output_start_y_tile + input_tile_height * self.config['network_g']['upscale'] # put tile into output image sr_images[:, :, output_start_y:output_end_y, output_start_x:output_end_x] = self.output_tile[:, :, output_start_y_tile:output_end_y_tile, output_start_x_tile:output_end_x_tile] del self.input_tile, self.output_tile, tile_batch, tile_list, output_tile_batch, output_tiles gc.collect() torch.cuda.empty_cache() return sr_images def train_model(self): if self.wandb_mode: wandb.init(project='HAT-for-image-sr', resume='allow', config= self.config, id=self.run_id) wandb.watch(self.model) if self.train_model_continue: epoch_lst = range(self.start_epoch, self.num_epochs) else: epoch_lst = range(self.num_epochs) for epoch in epoch_lst: start1 = time.time() # ------------------- TRAIN ------------------- if self.save_temp_model: self.load_network('temp_model_checkpoint.pth') self.model.train() train_epoch_loss = 0 stop = 0 for hr_images, lr_images in tqdm(self.train_dataloader, desc=f'Epoch {epoch+1}/{self.num_epochs}'): if stop == self.STOP: break stop+=1 loss = self.train_step(lr_images, hr_images) train_epoch_loss += loss if self.wandb_mode: wandb.log({ 'batch_loss': loss, }) if self.wandb_mode: wandb.log({ 'learning_rate': self.optimizer.param_groups[0]['lr'] }) print("Learning Rate is:", self.optimizer.param_groups[0]['lr']) self.scheduler.step() if self.save_temp_model: self.save_network(epoch, train_epoch_loss, 0, 'temp_model_checkpoint.pth') print(self.scheduler.state_dict()) self.del_model() del hr_images del lr_images gc.collect() train_epoch_loss /= len(self.train_dataloader) end1 = time.time() # ------------------- VALID ------------------- start2 = time.time() if self.save_temp_model: self.load_network('temp_model_checkpoint.pth') self.model.eval() with torch.no_grad(): valid_epoch_loss = 0 stop = 0 for hr_images, lr_images in tqdm(self.valid_dataloader, desc=f'Epoch {epoch+1}/{self.num_epochs}'): if stop == self.STOP: break stop+=1 loss = self.valid_step(lr_images, hr_images) valid_epoch_loss += loss valid_epoch_loss /= len(self.valid_dataloader) end2 = time.time() # ------------------- LOG ------------------- if self.wandb_mode: wandb.log({ 'train_loss': train_epoch_loss, 'valid_loss': valid_epoch_loss, }) # ------------------- VERBOSE ------------------- print(f'Epoch {epoch+1}/{self.num_epochs} | Train Loss: {train_epoch_loss:.4f} | Valid Loss: {valid_epoch_loss:.4f} | Time train: {end1-start1:.2f}s | Time valid: {end2-start2:.2f}s') # ------------------- CHECKPOINT ------------------- self.save_network(epoch, train_epoch_loss, valid_epoch_loss, 'model_checkpoint_latest.pth') if valid_epoch_loss < self.last_valid_loss: self.last_valid_loss = valid_epoch_loss self.save_network(epoch, train_epoch_loss, valid_epoch_loss, 'model_checkpoint_best.pth') print("New best checkpoint saved!") if self.save_temp_model: self.del_model() del hr_images del lr_images gc.collect() if self.wandb_mode: wandb.finish() def inference(self, lr_image, hr_image): """ - lr_image: torch.Tensor 3D Tensor (C, H, W) - hr_image: torch.Tesnor 3D Tensor (C, H, W). This parameter is optional, for comparing the model output and the ground-truth high-res image. If used solely for inference, skip this. Default is None/ """ lr_image = lr_image.unsqueeze(0).to(self.device) self.for_inference = True with torch.no_grad(): sr_image = self.tile_valid(lr_image) lr_image = lr_image.squeeze(0) sr_image = sr_image.squeeze(0) print(">> Size of low-res image:", lr_image.size()) print(">> Size of super-res image:", sr_image.size()) if hr_image != None: print(">> Size of high-res image:", hr_image.size()) if hr_image != None: fig, axes = plt.subplots(1, 3, figsize=(10, 6)) axes[0].imshow(lr_image.cpu().detach().permute((1, 2, 0))) axes[0].set_title('Low Resolution') axes[1].imshow(sr_image.cpu().detach().permute((1, 2, 0))) axes[1].set_title('Super Resolution') axes[2].imshow(hr_image.cpu().detach().permute((1, 2, 0))) axes[2].set_title('High Resolution') for ax in axes.flat: ax.axis('off') else: fig, axes = plt.subplots(1, 2, figsize=(10, 6)) axes[0].imshow(lr_image.cpu().detach().permute((1, 2, 0))) axes[0].set_title('Low Resolution') axes[1].imshow(sr_image.cpu().detach().permute((1, 2, 0))) axes[1].set_title('Super Resolution') for ax in axes.flat: ax.axis('off') plt.tight_layout() plt.show() return sr_image class TestDataset(Dataset): def __init__(self, lr_images_path): super(TestDataset, self).__init__() # hr_images_list = os.listdir(hr_images_path) self.lr_images_path = lr_images_path def __getitem__(self, idx): lr_image = Image.open(self.lr_image_path) lr_image = transforms.functional.to_tensor(lr_image) return lr_image if __name__ == "__main__": import os import sys # Getting to the Lambda directory sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), "../../")) image_path = "images/img_003_SRF_4_LR.png" infer_dataset = TestDataset(images_path=image_path) # hat = Network(run_id="hat-for-image-sr-" + str(int(1704006834)),config = config, wandb_mode = False, save_temp_model = True, train_model_continue = False) # STOP = 2 # num_params = sum(p.numel() for p in hat.model.parameters() if p.requires_grad) # print("Number of learnable parameters: ", num_params) # ---------- LOAD FROM LATEST CHECKPOINT ---------- # gc.collect() torch.cuda.empty_cache() hat = Network() hat.load_network(output) num_params = sum(p.numel() for p in hat.model.parameters() if p.requires_grad) print("Number of learnable parameters: ", num_params) image = image.squeeze(0) hat.inference(lr_image)