Delete testing_utils.py
Browse files- testing_utils.py +0 -210
testing_utils.py
DELETED
|
@@ -1,210 +0,0 @@
|
|
| 1 |
-
import argparse
|
| 2 |
-
import json
|
| 3 |
-
from PIL import Image
|
| 4 |
-
from torchvision import transforms
|
| 5 |
-
import torch.nn.functional as F
|
| 6 |
-
from glob import glob
|
| 7 |
-
|
| 8 |
-
import cv2
|
| 9 |
-
import math
|
| 10 |
-
import numpy as np
|
| 11 |
-
import os
|
| 12 |
-
import os.path as osp
|
| 13 |
-
import random
|
| 14 |
-
import time
|
| 15 |
-
import torch
|
| 16 |
-
from pathlib import Path
|
| 17 |
-
from torch.utils import data as data
|
| 18 |
-
|
| 19 |
-
from basicsr.utils import DiffJPEG, USMSharp
|
| 20 |
-
from basicsr.utils.img_process_util import filter2D
|
| 21 |
-
from basicsr.data.transforms import paired_random_crop, triplet_random_crop
|
| 22 |
-
from basicsr.data.degradations import random_add_gaussian_noise_pt, random_add_poisson_noise_pt, random_add_speckle_noise_pt, random_add_saltpepper_noise_pt, bivariate_Gaussian
|
| 23 |
-
|
| 24 |
-
from basicsr.data.degradations import circular_lowpass_kernel, random_mixed_kernels
|
| 25 |
-
from basicsr.data.transforms import augment
|
| 26 |
-
from basicsr.utils import FileClient, get_root_logger, imfrombytes, img2tensor
|
| 27 |
-
from basicsr.utils.registry import DATASET_REGISTRY
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
def parse_args_paired_testing(input_args=None):
|
| 31 |
-
"""
|
| 32 |
-
Parses command-line arguments used for configuring an paired session (pix2pix-Turbo).
|
| 33 |
-
This function sets up an argument parser to handle various training options.
|
| 34 |
-
|
| 35 |
-
Returns:
|
| 36 |
-
argparse.Namespace: The parsed command-line arguments.
|
| 37 |
-
"""
|
| 38 |
-
parser = argparse.ArgumentParser()
|
| 39 |
-
parser.add_argument("--ref_path", type=str, default=None,)
|
| 40 |
-
parser.add_argument("--base_config", default="./configs/sr_test.yaml", type=str)
|
| 41 |
-
parser.add_argument("--tracker_project_name", type=str, default="train_pix2pix_turbo", help="The name of the wandb project to log to.")
|
| 42 |
-
|
| 43 |
-
# details about the model architecture
|
| 44 |
-
parser.add_argument("--sd_path")
|
| 45 |
-
parser.add_argument("--de_net_path")
|
| 46 |
-
parser.add_argument("--pretrained_path", type=str, default=None,)
|
| 47 |
-
parser.add_argument("--revision", type=str, default=None,)
|
| 48 |
-
parser.add_argument("--variant", type=str, default=None,)
|
| 49 |
-
parser.add_argument("--tokenizer_name", type=str, default=None)
|
| 50 |
-
parser.add_argument("--lora_rank_unet", default=32, type=int)
|
| 51 |
-
parser.add_argument("--lora_rank_vae", default=16, type=int)
|
| 52 |
-
|
| 53 |
-
parser.add_argument("--scale", type=int, default=4, help="Scale factor for SR.")
|
| 54 |
-
parser.add_argument("--chop_size", type=int, default=128, choices=[512, 256, 128], help="Chopping forward.")
|
| 55 |
-
parser.add_argument("--chop_stride", type=int, default=96, help="Chopping stride.")
|
| 56 |
-
parser.add_argument("--padding_offset", type=int, default=32, help="padding offset.")
|
| 57 |
-
|
| 58 |
-
parser.add_argument("--vae_decoder_tiled_size", type=int, default=224)
|
| 59 |
-
parser.add_argument("--vae_encoder_tiled_size", type=int, default=1024)
|
| 60 |
-
parser.add_argument("--latent_tiled_size", type=int, default=96)
|
| 61 |
-
parser.add_argument("--latent_tiled_overlap", type=int, default=32)
|
| 62 |
-
|
| 63 |
-
parser.add_argument("--align_method", type=str, default="wavelet")
|
| 64 |
-
|
| 65 |
-
parser.add_argument("--pos_prompt", type=str, default="A high-resolution, 8K, ultra-realistic image with sharp focus, vibrant colors, and natural lighting.")
|
| 66 |
-
parser.add_argument("--neg_prompt", type=str, default="oil painting, cartoon, blur, dirty, messy, low quality, deformation, low resolution, oversmooth")
|
| 67 |
-
|
| 68 |
-
# training details
|
| 69 |
-
parser.add_argument("--output_dir", required=True)
|
| 70 |
-
parser.add_argument("--cache_dir", default=None,)
|
| 71 |
-
parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
|
| 72 |
-
parser.add_argument("--resolution", type=int, default=512,)
|
| 73 |
-
parser.add_argument("--checkpointing_steps", type=int, default=500,)
|
| 74 |
-
parser.add_argument("--gradient_accumulation_steps", type=int, default=1, help="Number of updates steps to accumulate before performing a backward/update pass.",)
|
| 75 |
-
parser.add_argument("--gradient_checkpointing", action="store_true",)
|
| 76 |
-
|
| 77 |
-
parser.add_argument("--dataloader_num_workers", type=int, default=0,)
|
| 78 |
-
parser.add_argument("--allow_tf32", action="store_true",
|
| 79 |
-
help=(
|
| 80 |
-
"Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
|
| 81 |
-
" https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
|
| 82 |
-
),
|
| 83 |
-
)
|
| 84 |
-
parser.add_argument("--report_to", type=str, default="wandb",
|
| 85 |
-
help=(
|
| 86 |
-
'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
|
| 87 |
-
' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
|
| 88 |
-
),
|
| 89 |
-
)
|
| 90 |
-
parser.add_argument("--mixed_precision", type=str, default=None, choices=["no", "fp16", "bf16"],)
|
| 91 |
-
parser.add_argument("--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers.")
|
| 92 |
-
parser.add_argument("--set_grads_to_none", action="store_true",)
|
| 93 |
-
|
| 94 |
-
parser.add_argument('--world_size', default=1, type=int,
|
| 95 |
-
help='number of distributed processes')
|
| 96 |
-
parser.add_argument('--local_rank', default=-1, type=int)
|
| 97 |
-
parser.add_argument('--dist_url', default='env://',
|
| 98 |
-
help='url used to set up distributed training')
|
| 99 |
-
|
| 100 |
-
if input_args is not None:
|
| 101 |
-
args = parser.parse_args(input_args)
|
| 102 |
-
else:
|
| 103 |
-
args = parser.parse_args()
|
| 104 |
-
|
| 105 |
-
return args
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
class PlainDataset(data.Dataset):
|
| 109 |
-
"""Modified dataset based on the dataset used for Real-ESRGAN model:
|
| 110 |
-
Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data.
|
| 111 |
-
|
| 112 |
-
It loads gt (Ground-Truth) images, and augments them.
|
| 113 |
-
It also generates blur kernels and sinc kernels for generating low-quality images.
|
| 114 |
-
Note that the low-quality images are processed in tensors on GPUS for faster processing.
|
| 115 |
-
|
| 116 |
-
Args:
|
| 117 |
-
opt (dict): Config for train datasets. It contains the following keys:
|
| 118 |
-
dataroot_gt (str): Data root path for gt.
|
| 119 |
-
meta_info (str): Path for meta information file.
|
| 120 |
-
io_backend (dict): IO backend type and other kwarg.
|
| 121 |
-
use_hflip (bool): Use horizontal flips.
|
| 122 |
-
use_rot (bool): Use rotation (use vertical flip and transposing h and w for implementation).
|
| 123 |
-
Please see more options in the codes.
|
| 124 |
-
"""
|
| 125 |
-
|
| 126 |
-
def __init__(self, opt):
|
| 127 |
-
super(PlainDataset, self).__init__()
|
| 128 |
-
self.opt = opt
|
| 129 |
-
self.file_client = None
|
| 130 |
-
self.io_backend_opt = opt['io_backend']
|
| 131 |
-
|
| 132 |
-
if 'image_type' not in opt:
|
| 133 |
-
opt['image_type'] = 'png'
|
| 134 |
-
|
| 135 |
-
# support multiple type of data: file path and meta data, remove support of lmdb
|
| 136 |
-
self.lr_paths = []
|
| 137 |
-
if 'lr_path' in opt:
|
| 138 |
-
if isinstance(opt['lr_path'], str):
|
| 139 |
-
self.lr_paths.extend(sorted(
|
| 140 |
-
[str(x) for x in Path(opt['lr_path']).glob('*.png')] +
|
| 141 |
-
[str(x) for x in Path(opt['lr_path']).glob('*.jpg')] +
|
| 142 |
-
[str(x) for x in Path(opt['lr_path']).glob('*.jpeg')]
|
| 143 |
-
))
|
| 144 |
-
else:
|
| 145 |
-
self.lr_paths.extend(sorted([str(x) for x in Path(opt['lr_path'][0]).glob('*.'+opt['image_type'])]))
|
| 146 |
-
if len(opt['lr_path']) > 1:
|
| 147 |
-
for i in range(len(opt['lr_path'])-1):
|
| 148 |
-
self.lr_paths.extend(sorted([str(x) for x in Path(opt['lr_path'][i+1]).glob('*.'+opt['image_type'])]))
|
| 149 |
-
|
| 150 |
-
def __getitem__(self, index):
|
| 151 |
-
if self.file_client is None:
|
| 152 |
-
self.file_client = FileClient(self.io_backend_opt.pop('type'), **self.io_backend_opt)
|
| 153 |
-
|
| 154 |
-
# -------------------------------- Load gt images -------------------------------- #
|
| 155 |
-
# Shape: (h, w, c); channel order: BGR; image range: [0, 1], float32.
|
| 156 |
-
lr_path = self.lr_paths[index]
|
| 157 |
-
|
| 158 |
-
# avoid errors caused by high latency in reading files
|
| 159 |
-
retry = 3
|
| 160 |
-
while retry > 0:
|
| 161 |
-
try:
|
| 162 |
-
lr_img_bytes = self.file_client.get(lr_path, 'gt')
|
| 163 |
-
except (IOError, OSError) as e:
|
| 164 |
-
# logger = get_root_logger()
|
| 165 |
-
# logger.warn(f'File client error: {e}, remaining retry times: {retry - 1}')
|
| 166 |
-
# change another file to read
|
| 167 |
-
index = random.randint(0, self.__len__()-1)
|
| 168 |
-
lr_path = self.lr_paths[index]
|
| 169 |
-
time.sleep(1) # sleep 1s for occasional server congestion
|
| 170 |
-
else:
|
| 171 |
-
break
|
| 172 |
-
finally:
|
| 173 |
-
retry -= 1
|
| 174 |
-
|
| 175 |
-
img_lr = imfrombytes(lr_img_bytes, float32=True)
|
| 176 |
-
|
| 177 |
-
# BGR to RGB, HWC to CHW, numpy to tensor
|
| 178 |
-
img_lr = img2tensor([img_lr], bgr2rgb=True, float32=True)[0]
|
| 179 |
-
|
| 180 |
-
return_d = {'lr': img_lr, 'lr_path': lr_path}
|
| 181 |
-
return return_d
|
| 182 |
-
|
| 183 |
-
def __len__(self):
|
| 184 |
-
return len(self.lr_paths)
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
def lr_proc(config, batch, device):
|
| 188 |
-
im_lr = batch['lr'].cuda()
|
| 189 |
-
im_lr = im_lr.to(memory_format=torch.contiguous_format).float()
|
| 190 |
-
|
| 191 |
-
ori_lr = im_lr
|
| 192 |
-
|
| 193 |
-
im_lr = F.interpolate(
|
| 194 |
-
im_lr,
|
| 195 |
-
size=(im_lr.size(-2) * config.sf,
|
| 196 |
-
im_lr.size(-1) * config.sf),
|
| 197 |
-
mode='bicubic',
|
| 198 |
-
)
|
| 199 |
-
|
| 200 |
-
im_lr = im_lr.contiguous()
|
| 201 |
-
im_lr = im_lr * 2 - 1.0
|
| 202 |
-
im_lr = torch.clamp(im_lr, -1.0, 1.0)
|
| 203 |
-
|
| 204 |
-
ori_h, ori_w = im_lr.size(-2), im_lr.size(-1)
|
| 205 |
-
|
| 206 |
-
pad_h = (math.ceil(ori_h / 64)) * 64 - ori_h
|
| 207 |
-
pad_w = (math.ceil(ori_w / 64)) * 64 - ori_w
|
| 208 |
-
im_lr = F.pad(im_lr, pad=(0, pad_w, 0, pad_h), mode='reflect')
|
| 209 |
-
|
| 210 |
-
return im_lr.to(device), ori_lr.to(device), (ori_h, ori_w)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|