Managed to get the training colab notebook to work! Was battling with the code for over 24 hrs.

For some reason the training will only go up to "Epoch: 4 Iter: 2500" (15 mins or so) then it stops, and I would like to train it for longer if possible, is there some setting I'm missing?

i have a problem related to Wav2Lip-HQ pretraining. when i run Finetune ESRGAN i get this ouput:

Traceback (most recent call last):
File "basicsr/train.py", line 17, in
from basicsr.utils.options import dict2str, parse
ImportError: cannot import name 'parse' from 'basicsr.utils.options' (/usr/local/lib/python3.7/dist-packages/basicsr/utils/options.py)

Is there any solution?

thanks for your interesting work !
I can not get the same results you mentioned in README?
Is the input format/ language important?

Are there any tutorials?

脸色灰白 人丑，嘴巴 发光

The colab pretraining notebook is not updated. There's a lot of bugs on the code (related to paths and inexistent files)
https://colab.research.google.com/drive/1IUGYn-fMRbjH2IyYoAn5VKSzEkaXyP2s

I want to know how to prepare my own customized data for finetuning part (data is in .mp4 format). I would like to train the model using my own video to improve the resolution. May you explain the procedure in details?

Thank you.

FileNotFoundError: [Errno 2] No such file or directory: 'checkpoints/esrgan_max.pth'

Using cuda for inference.
(80, 2612)
Length of mel chunks: 813
  0%|                                                                                             | 0/7 [00:00<?, ?it/s]Reading video frames from start...
  0%|                                                                                             | 0/7 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "inference.py", line 355, in <module>
    main()
  File "inference.py", line 292, in main
    for i, (img_batch, mel_batch, frames, coords) in enumerate(tqdm(gen,
  File "/home/lizhaohui/miniconda3/envs/wav2lip/lib/python3.8/site-packages/tqdm/std.py", line 1195, in __iter__
    for obj in iterable:
  File "inference.py", line 156, in datagen
    face, coords = face_detect([frame_to_save])[0]
  File "inference.py", line 93, in face_detect
    detector = face_detection.FaceAlignment(face_detection.LandmarksType._2D,
  File "/home/lizhaohui/wav2lip-hq/face_detection/api.py", line 62, in __init__
    self.face_detector = face_detector_module.FaceDetector(device=device, verbose=verbose)
  File "/home/lizhaohui/wav2lip-hq/face_detection/detection/sfd/sfd_detector.py", line 24, in __init__
    model_weights = torch.load(path_to_detector)
  File "/home/lizhaohui/miniconda3/envs/wav2lip/lib/python3.8/site-packages/torch/serialization.py", line 713, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/home/lizhaohui/miniconda3/envs/wav2lip/lib/python3.8/site-packages/torch/serialization.py", line 920, in _legacy_load
    magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, '!'.

The above error occurred in my operation

测试案例中的嘴唇效果很差，感觉在乱动，是否重新训练会好一些。

There doesn't appear to a working link to relevant face_segmentation.pth file? Please can update the link.

I am trying to do inferencing but to some reason it's explicitly asking for GPU.

my script is:
python inference.py --checkpoint_path wav2lip_gan.pth --segmentation_path s3fd-619a316812.pth --sr_path esrgan_yunying.pth --face chroma_video_1.mp4 --audio wave_01.mp3 --outfile result.mp4

and the error is:

File "inference.py", line 296, in main
    seg_net = init_parser(args.segmentation_path)
...................................
...................................
File "/home/ubuntu/anaconda3/envs/myenv/lib/python3.7/site-packages/torch/cuda/__init__.py", line 172, in _lazy_init
    torch._C._cuda_init()
RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx

on face_parsing/swap.py file,
under this function, is calling net.cuda(), how do i change it to cpu?

def init_parser(pth_path):
    n_classes = 19
    net = BiSeNet(n_classes=n_classes)
    net.cuda()
    net.load_state_dict(torch.load(pth_path))
    net.eval()
    return net

Wheni Run, Run Wav2Lip with frame saving., i get a error check the photos below, do i need to run Wav2Lip-HQ pretraining.ipynb, first is that the step 1, then i go to run the 2 step is that correct, Wav2Lip-HQ inference.ipynb,

*

the result like this

Why it looks so different from you：

Hello, author, could you provide the document:“net_g_67500.pth”?

I use 8G gpu, but the process is killed.

Disable distributed.
2021-07-26 08:47:50,661 INFO:
____ _ _____ ____
/ __ ) ____ _ _____ ()____/ / / __
/ __ |/ __ `// // // /_ \ / // /
/ // // // /( )/ // / / // _, _/
// _,///// ___///// ||
______ __ __ __ __
/ / ____ / / / / __ __ _____ / / / /
/ / __ / __ \ / __ \ / __ / / / / / / // // /// / /
/ // // // // // // // / / // // // / / /< //
_/ _/ _/ _/ //_/ __///|| (_)

Version Information:
BasicSR: 1.3.3.10
PyTorch: 1.9.0+cu102
TorchVision: 0.10.0+cu102
2021-07-26 08:47:50,661 INFO:
name: 001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb
model_type: ESRGANModel
scale: 4
num_gpu: 1
manual_seed: 0
datasets:[
train:[
name: face_dataset
type: PairedImageDataset
dataroot_gt: data/hq
dataroot_lq: data/lq
filename_tmpl: {}
io_backend:[
type: disk
]
gt_size: 384
use_flip: True
use_rot: True
use_shuffle: True
num_worker_per_gpu: 1
batch_size_per_gpu: 4
dataset_enlarge_ratio: 1
prefetch_mode: None
phase: train
scale: 4
]
]
network_g:[
type: RRDBNet
num_in_ch: 3
num_out_ch: 3
num_feat: 64
num_block: 23
]
network_d:[
type: VGGStyleDiscriminator128
num_in_ch: 3
num_feat: 64
]
path:[
pretrain_network_g: None
strict_load_g: True
resume_state: checkpoints/pretrained.state
experiments_root: /content/wav2lip-hq/experiments/001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb
models: /content/wav2lip-hq/experiments/001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb/models
training_states: /content/wav2lip-hq/experiments/001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb/training_states
log: /content/wav2lip-hq/experiments/001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb
visualization: /content/wav2lip-hq/experiments/001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb/visualization
]
train:[
optim_g:[
type: Adam
lr: 0.0001
weight_decay: 0
betas: [0.9, 0.99]
]
optim_d:[
type: Adam
lr: 0.0001
weight_decay: 0
betas: [0.9, 0.99]
]
scheduler:[
type: MultiStepLR
milestones: [50000, 100000, 200000, 300000]
gamma: 0.5
]
total_iter: 150000
warmup_iter: -1
pixel_opt:[
type: L1Loss
loss_weight: 0.01
reduction: mean
]
perceptual_opt:[
type: PerceptualLoss
layer_weights:[
conv5_4: 1
]
vgg_type: vgg19
use_input_norm: True
range_norm: False
perceptual_weight: 1.0
style_weight: 0
criterion: l1
]
gan_opt:[
type: GANLoss
gan_type: vanilla
real_label_val: 1.0
fake_label_val: 0.0
loss_weight: 0.005
]
net_d_iters: 1
net_d_init_iters: 0
]
val:[
val_freq: 2500.0
save_img: True
metrics:[
psnr:[
type: calculate_psnr
crop_border: 4
test_y_channel: False
]
]
]
logger:[
print_freq: 100
save_checkpoint_freq: 2500.0
use_tb_logger: True
wandb:[
project: None
resume_id: None
]
]
dist_params:[
backend: nccl
port: 29500
]
is_train: True
dist: False
rank: 0
world_size: 1

2021-07-26 08:47:50.932429: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-07-26 08:47:52,184 INFO: Dataset [PairedImageDataset] - face_dataset is built.
2021-07-26 08:47:52,184 INFO: Training statistics:
Number of train images: 1730
Dataset enlarge ratio: 1
Batch size per gpu: 4
World size (gpu number): 1
Require iter number per epoch: 433
Total epochs: 346; iters: 150000.
Set pretrain_network_g to /content/wav2lip-hq/experiments/001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb/models/net_g_67500.pth
Set pretrain_network_d to /content/wav2lip-hq/experiments/001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb/models/net_d_67500.pth
2021-07-26 08:47:52,497 INFO: Network [RRDBNet] is created.
2021-07-26 08:47:52,553 INFO: Network: RRDBNet, with parameters: 16,697,987
2021-07-26 08:47:52,553 INFO: RRDBNet(
(conv_first): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(body): Sequential(
(0): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(1): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(2): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(3): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(4): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(5): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(6): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(7): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(8): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(9): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(10): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(11): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(12): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(13): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(14): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(15): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(16): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(17): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(18): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(19): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(20): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(21): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
(22): RRDB(
(rdb1): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb2): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
(rdb3): ResidualDenseBlock(
(conv1): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv2): Conv2d(96, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv3): Conv2d(128, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv4): Conv2d(160, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv5): Conv2d(192, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
)
)
(conv_body): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv_up1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv_up2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv_hr): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(conv_last): Conv2d(64, 3, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(lrelu): LeakyReLU(negative_slope=0.2, inplace=True)
)
2021-07-26 08:47:52,616 INFO: Loading RRDBNet model from /content/wav2lip-hq/experiments/001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb/models/net_g_67500.pth.
Traceback (most recent call last):
File "basicsr/train.py", line 221, in
train_pipeline(root_path)
File "basicsr/train.py", line 138, in train_pipeline
model = build_model(opt)
File "/usr/local/lib/python3.7/dist-packages/basicsr/models/init.py", line 27, in build_model
model = MODEL_REGISTRY.get(opt['model_type'])(opt)
File "/usr/local/lib/python3.7/dist-packages/basicsr/models/sr_model.py", line 29, in init
self.load_network(self.net_g, load_path, self.opt['path'].get('strict_load_g', True))
File "/usr/local/lib/python3.7/dist-packages/basicsr/models/base_model.py", line 265, in load_network
load_net = torch.load(load_path, map_location=lambda storage, loc: storage)
File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 608, in load
return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 794, in _legacy_load
deserialized_objects[key]._set_from_file(f, offset, f_should_read_directly)
RuntimeError: storage has wrong size: expected -4754300013468754645 got 32

What is the difference between ersgan_max and ersgan_yunying?

RuntimeError: Error(s) in loading state_dict for VGGStyleDiscriminator128:
Missing key(s) in state_dict: "conv0_0.weight", "conv0_0.bias", "conv0_1.weight", "bn0_1.weight", "bn0_1.bias", "bn0_1.running_mean", "bn0_1.running_var", "conv1_0.weight", "bn1_0.weight", "bn1_0.bias", "bn1_0.running_mean", "bn1_0.running_var", "conv1_1.weight", "bn1_1.weight", "bn1_1.bias", "bn1_1.running_mean", "bn1_1.running_var", "conv2_0.weight", "bn2_0.weight", "bn2_0.bias", "bn2_0.running_mean", "bn2_0.running_var", "conv2_1.weight", "bn2_1.weight", "bn2_1.bias", "bn2_1.running_mean", "bn2_1.running_var", "conv3_0.weight", "bn3_0.weight", "bn3_0.bias", "bn3_0.running_mean", "bn3_0.running_var", "conv3_1.weight", "bn3_1.weight", "bn3_1.bias", "bn3_1.running_mean", "bn3_1.running_var", "conv4_0.weight", "bn4_0.weight", "bn4_0.bias", "bn4_0.running_mean", "bn4_0.running_var", "conv4_1.weight", "bn4_1.weight", "bn4_1.bias", "bn4_1.running_mean", "bn4_1.running_var", "linear1.weight", "linear1.bias", "linear2.weight", "linear2.bias".
Unexpected key(s) in state_dict: "conv_first.weight", "conv_first.bias", "body.0.rdb1.conv1.weight", "body.0.rdb1.conv1.bias", "body.0.rdb1.conv2.weight", "body.0.rdb1.conv2.bias", "body.0.rdb1.conv3.weight", "body.0.rdb1.conv3.bias", "body.0.rdb1.conv4.weight", "body.0.rdb1.conv4.bias", "body.0.rdb1.conv5.weight", "body.0.rdb1.conv5.bias", "body.0.rdb2.conv1.weight", "body.0.rdb2.conv1.bias", "body.0.rdb2.conv2.weight", "body.0.rdb2.conv2.bias", "body.0.rdb2.conv3.weight", "body.0.rdb2.conv3.bias", "body.0.rdb2.conv4.weight", "body.0.rdb2.conv4.bias", "body.0.rdb2.conv5.weight", "body.0.rdb2.conv5.bias", "body.0.rdb3.conv1.weight", "body.0.rdb3.conv1.bias", "body.0.rdb3.conv2.weight", "body.0.rdb3.conv2.bias", "body.0.rdb3.conv3.weight", "body.0.rdb3.conv3.bias", "body.0.rdb3.conv4.weight", "body.0.rdb3.conv4.bias", "body.0.rdb3.conv5.weight", "body.0.rdb3.conv5.bias", "body.1.rdb1.conv1.weight", "body.1.rdb1.conv1.bias", "body.1.rdb1.conv2.weight", "body.1.rdb1.conv2.bias", "body.1.rdb1.conv3.weight", "body.1.rdb1.conv3.bias", "body.1.rdb1.conv4.weight", "body.1.rdb1.conv4.bias", "body.1.rdb1.conv5.weight", "body.1.rdb1.conv5.bias", "body.1.rdb2.conv1.weight", "body.1.rdb2.conv1.bias", "body.1.rdb2.conv2.weight", "body.1.rdb2.conv2.bias", "body.1.rdb2.conv3.weight", "body.1.rdb2.conv3.bias", "body.1.rdb2.conv4.weight", "body.1.rdb2.conv4.bias", "body.1.rdb2.conv5.weight", "body.1.rdb2.conv5.bias", "body.1.rdb3.conv1.weight", "body.1.rdb3.conv1.bias", "body.1.rdb3.conv2.weight", "body.1.rdb3.conv2.bias", "body.1.rdb3.conv3.weight", "body.1.rdb3.conv3.bias", "body.1.rdb3.conv4.weight", "body.1.rdb3.conv4.bias", "body.1.rdb3.conv5.weight", "body.1.rdb3.conv5.bias", "body.2.rdb1.conv1.weight", "body.2.rdb1.conv1.bias", "body.2.rdb1.conv2.weight", "body.2.rdb1.conv2.bias", "body.2.rdb1.conv3.weight", "body.2.rdb1.conv3.bias", "body.2.rdb1.conv4.weight", "body.2.rdb1.conv4.bias", "body.2.rdb1.conv5.weight", "body.2.rdb1.conv5.bias", "body.2.rdb2.conv1.weight", "body.2.rdb2.conv1.bias", "body.2.rdb2.conv2.weight", "body.2.rdb2.conv2.bias", "body.2.rdb2.conv3.weight", "body.2.rdb2.conv3.bias", "body.2.rdb2.conv4.weight", "body.2.rdb2.conv4.bias", "body.2.rdb2.conv5.weight", "body.2.rdb2.conv5.bias", "body.2.rdb3.conv1.weight", "body.2.rdb3.conv1.bias", "body.2.rdb3.conv2.weight", "body.2.rdb3.conv2.bias", "body.2.rdb3.conv3.weight", "body.2.rdb3.conv3.bias", "body.2.rdb3.conv4.weight", "body.2.rdb3.conv4.bias", "body.2.rdb3.conv5.weight", "body.2.rdb3.conv5.bias", "body.3.rdb1.conv1.weight", "body.3.rdb1.conv1.bias", "body.3.rdb1.conv2.weight", "body.3.rdb1.conv2.bias", "body.3.rdb1.conv3.weight", "body.3.rdb1.conv3.bias", "body.3.rdb1.conv4.weight", "body.3.rdb1.conv4.bias", "body.3.rdb1.conv5.weight", "body.3.rdb1.conv5.bias", "body.3.rdb2.conv1.weight", "body.3.rdb2.conv1.bias", "body.3.rdb2.conv2.weight", "body.3.rdb2.conv2.bias", "body.3.rdb2.conv3.weight", "body.3.rdb2.conv3.bias", "body.3.rdb2.conv4.weight", "body.3.rdb2.conv4.bias", "body.3.rdb2.conv5.weight", "body.3.rdb2.conv5.bias", "body.3.rdb3.conv1.weight", "body.3.rdb3.conv1.bias", "body.3.rdb3.conv2.weight", "body.3.rdb3.conv2.bias", "body.3.rdb3.conv3.weight", "body.3.rdb3.conv3.bias", "body.3.rdb3.conv4.weight", "body.3.rdb3.conv4.bias", "body.3.rdb3.conv5.weight", "body.3.rdb3.conv5.bias", "body.4.rdb1.conv1.weight", "body.4.rdb1.conv1.bias", "body.4.rdb1.conv2.weight", "body.4.rdb1.conv2.bias", "body.4.rdb1.conv3.weight", "body.4.rdb1.conv3.bias", "body.4.rdb1.conv4.weight", "body.4.rdb1.conv4.bias", "body.4.rdb1.conv5.weight", "body.4.rdb1.conv5.bias", "body.4.rdb2.conv1.weight", "body.4.rdb2.conv1.bias", "body.4.rdb2.conv2.weight", "body.4.rdb2.conv2.bias", "body.4.rdb2.conv3.weight", "body.4.rdb2.conv3.bias", "body.4.rdb2.conv4.weight", "body.4.rdb2.conv4.bias", "body.4.rdb2.conv5.weight", "body.4.rdb2.conv5.bias", "body.4.rdb3.conv1.weight", "body.4.rdb3.conv1.bias", "body.4.rdb3.conv2.weight", "body.4.rdb3.conv2.bias", "body.4.rdb3.conv3.weight", "body.4.rdb3.conv3.bias", "body.4.rdb3.conv4.weight", "body.4.rdb3.conv4.bias", "body.4.rdb3.conv5.weight", "body.4.rdb3.conv5.bias", "body.5.rdb1.conv1.weight", "body.5.rdb1.conv1.bias", "body.5.rdb1.conv2.weight", "body.5.rdb1.conv2.bias", "body.5.rdb1.conv3.weight", "body.5.rdb1.conv3.bias", "body.5.rdb1.conv4.weight", "body.5.rdb1.conv4.bias", "body.5.rdb1.conv5.weight", "body.5.rdb1.conv5.bias", "body.5.rdb2.conv1.weight", "body.5.rdb2.conv1.bias", "body.5.rdb2.conv2.weight", "body.5.rdb2.conv2.bias", "body.5.rdb2.conv3.weight", "body.5.rdb2.conv3.bias", "body.5.rdb2.conv4.weight", "body.5.rdb2.conv4.bias", "body.5.rdb2.conv5.weight", "body.5.rdb2.conv5.bias", "body.5.rdb3.conv1.weight", "body.5.rdb3.conv1.bias", "body.5.rdb3.conv2.weight", "body.5.rdb3.conv2.bias", "body.5.rdb3.conv3.weight", "body.5.rdb3.conv3.bias", "body.5.rdb3.conv4.weight", "body.5.rdb3.conv4.bias", "body.5.rdb3.conv5.weight", "body.5.rdb3.conv5.bias", "body.6.rdb1.conv1.weight", "body.6.rdb1.conv1.bias", "body.6.rdb1.conv2.weight", "body.6.rdb1.conv2.bias", "body.6.rdb1.conv3.weight", "body.6.rdb1.conv3.bias", "body.6.rdb1.conv4.weight", "body.6.rdb1.conv4.bias", "body.6.rdb1.conv5.weight", "body.6.rdb1.conv5.bias", "body.6.rdb2.conv1.weight", "body.6.rdb2.conv1.bias", "body.6.rdb2.conv2.weight", "body.6.rdb2.conv2.bias", "body.6.rdb2.conv3.weight", "body.6.rdb2.conv3.bias", "body.6.rdb2.conv4.weight", "body.6.rdb2.conv4.bias", "body.6.rdb2.conv5.weight", "body.6.rdb2.conv5.bias", "body.6.rdb3.conv1.weight", "body.6.rdb3.conv1.bias", "body.6.rdb3.conv2.weight", "body.6.rdb3.conv2.bias", "body.6.rdb3.conv3.weight", "body.6.rdb3.conv3.bias", "body.6.rdb3.conv4.weight", "body.6.rdb3.conv4.bias", "body.6.rdb3.conv5.weight", "body.6.rdb3.conv5.bias", "body.7.rdb1.conv1.weight", "body.7.rdb1.conv1.bias", "body.7.rdb1.conv2.weight", "body.7.rdb1.conv2.bias", "body.7.rdb1.conv3.weight", "body.7.rdb1.conv3.bias", "body.7.rdb1.conv4.weight", "body.7.rdb1.conv4.bias", "body.7.rdb1.conv5.weight", "body.7.rdb1.conv5.bias", "body.7.rdb2.conv1.weight", "body.7.rdb2.conv1.bias", "body.7.rdb2.conv2.weight", "body.7.rdb2.conv2.bias", "body.7.rdb2.conv3.weight", "body.7.rdb2.conv3.bias", "body.7.rdb2.conv4.weight", "body.7.rdb2.conv4.bias", "body.7.rdb2.conv5.weight", "body.7.rdb2.conv5.bias", "body.7.rdb3.conv1.weight", "body.7.rdb3.conv1.bias", "body.7.rdb3.conv2.weight", "body.7.rdb3.conv2.bias", "body.7.rdb3.conv3.weight", "body.7.rdb3.conv3.bias", "body.7.rdb3.conv4.weight", "body.7.rdb3.conv4.bias", "body.7.rdb3.conv5.weight", "body.7.rdb3.conv5.bias", "body.8.rdb1.conv1.weight", "body.8.rdb1.conv1.bias", "body.8.rdb1.conv2.weight", "body.8.rdb1.conv2.bias", "body.8.rdb1.conv3.weight", "body.8.rdb1.conv3.bias", "body.8.rdb1.conv4.weight", "body.8.rdb1.conv4.bias", "body.8.rdb1.conv5.weight", "body.8.rdb1.conv5.bias", "body.8.rdb2.conv1.weight", "body.8.rdb2.conv1.bias", "body.8.rdb2.conv2.weight", "body.8.rdb2.conv2.bias", "body.8.rdb2.conv3.weight", "body.8.rdb2.conv3.bias", "body.8.rdb2.conv4.weight", "body.8.rdb2.conv4.bias", "body.8.rdb2.conv5.weight", "body.8.rdb2.conv5.bias", "body.8.rdb3.conv1.weight", "body.8.rdb3.conv1.bias", "body.8.rdb3.conv2.weight", "body.8.rdb3.conv2.bias", "body.8.rdb3.conv3.weight", "body.8.rdb3.conv3.bias", "body.8.rdb3.conv4.weight", "body.8.rdb3.conv4.bias", "body.8.rdb3.conv5.weight", "body.8.rdb3.conv5.bias", "body.9.rdb1.conv1.weight", "body.9.rdb1.conv1.bias", "body.9.rdb1.conv2.weight", "body.9.rdb1.conv2.bias", "body.9.rdb1.conv3.weight", "body.9.rdb1.conv3.bias", "body.9.rdb1.conv4.weight", "body.9.rdb1.conv4.bias", "body.9.rdb1.conv5.weight", "body.9.rdb1.conv5.bias", "body.9.rdb2.conv1.weight", "body.9.rdb2.conv1.bias", "body.9.rdb2.conv2.weight", "body.9.rdb2.conv2.bias", "body.9.rdb2.conv3.weight", "body.9.rdb2.conv3.bias", "body.9.rdb2.conv4.weight", "body.9.rdb2.conv4.bias", "body.9.rdb2.conv5.weight", "body.9.rdb2.conv5.bias", "body.9.rdb3.conv1.weight", "body.9.rdb3.conv1.bias", "body.9.rdb3.conv2.weight", "body.9.rdb3.conv2.bias", "body.9.rdb3.conv3.weight", "body.9.rdb3.conv3.bias", "body.9.rdb3.conv4.weight", "body.9.rdb3.conv4.bias", "body.9.rdb3.conv5.weight", "body.9.rdb3.conv5.bias", "body.10.rdb1.conv1.weight", "body.10.rdb1.conv1.bias", "body.10.rdb1.conv2.weight", "body.10.rdb1.conv2.bias", "body.10.rdb1.conv3.weight", "body.10.rdb1.conv3.bias", "body.10.rdb1.conv4.weight", "body.10.rdb1.conv4.bias", "body.10.rdb1.conv5.weight", "body.10.rdb1.conv5.bias", "body.10.rdb2.conv1.weight", "body.10.rdb2.conv1.bias", "body.10.rdb2.conv2.weight", "body.10.rdb2.conv2.bias", "body.10.rdb2.conv3.weight", "body.10.rdb2.conv3.bias", "body.10.rdb2.conv4.weight", "body.10.rdb2.conv4.bias", "body.10.rdb2.conv5.weight", "body.10.rdb2.conv5.bias", "body.10.rdb3.conv1.weight", "body.10.rdb3.conv1.bias", "body.10.rdb3.conv2.weight", "body.10.rdb3.conv2.bias", "body.10.rdb3.conv3.weight", "body.10.rdb3.conv3.bias", "body.10.rdb3.conv4.weight", "body.10.rdb3.conv4.bias", "body.10.rdb3.conv5.weight", "body.10.rdb3.conv5.bias", "body.11.rdb1.conv1.weight", "body.11.rdb1.conv1.bias", "body.11.rdb1.conv2.weight", "body.11.rdb1.conv2.bias", "body.11.rdb1.conv3.weight", "body.11.rdb1.conv3.bias", "body.11.rdb1.conv4.weight", "body.11.rdb1.conv4.bias", "body.11.rdb1.conv5.weight", "body.11.rdb1.conv5.bias", "body.11.rdb2.conv1.weight", "body.11.rdb2.conv1.bias", "body.11.rdb2.conv2.weight", "body.11.rdb2.conv2.bias", "body.11.rdb2.conv3.weight", "body.11.rdb2.conv3.bias", "body.11.rdb2.conv4.weight", "body.11.rdb2.conv4.bias", "body.11.rdb2.conv5.weight", "body.11.rdb2.conv5.bias", "body.11.rdb3.conv1.weight", "body.11.rdb3.conv1.bias", "body.11.rdb3.conv2.weight", "body.11.rdb3.conv2.bias", "body.11.rdb3.conv3.weight", "body.11.rdb3.conv3.bias", "body.11.rdb3.conv4.weight", "body.11.rdb3.conv4.bias", "body.11.rdb3.conv5.weight", "body.11.rdb3.conv5.bias", "body.12.rdb1.conv1.weight", "body.12.rdb1.conv1.bias", "body.12.rdb1.conv2.weight", "body.12.rdb1.conv2.bias", "body.12.rdb1.conv3.weight", "body.12.rdb1.conv3.bias", "body.12.rdb1.conv4.weight", "body.12.rdb1.conv4.bias", "body.12.rdb1.conv5.weight", "body.12.rdb1.conv5.bias", "body.12.rdb2.conv1.weight", "body.12.rdb2.conv1.bias", "body.12.rdb2.conv2.weight", "body.12.rdb2.conv2.bias", "body.12.rdb2.conv3.weight", "body.12.rdb2.conv3.bias", "body.12.rdb2.conv4.weight", "body.12.rdb2.conv4.bias", "body.12.rdb2.conv5.weight", "body.12.rdb2.conv5.bias", "body.12.rdb3.conv1.weight", "body.12.rdb3.conv1.bias", "body.12.rdb3.conv2.weight", "body.12.rdb3.conv2.bias", "body.12.rdb3.conv3.weight", "body.12.rdb3.conv3.bias", "body.12.rdb3.conv4.weight", "body.12.rdb3.conv4.bias", "body.12.rdb3.conv5.weight", "body.12.rdb3.conv5.bias", "body.13.rdb1.conv1.weight", "body.13.rdb1.conv1.bias", "body.13.rdb1.conv2.weight", "body.13.rdb1.conv2.bias", "body.13.rdb1.conv3.weight", "body.13.rdb1.conv3.bias", "body.13.rdb1.conv4.weight", "body.13.rdb1.conv4.bias", "body.13.rdb1.conv5.weight", "body.13.rdb1.conv5.bias", "body.13.rdb2.conv1.weight", "body.13.rdb2.conv1.bias", "body.13.rdb2.conv2.weight", "body.13.rdb2.conv2.bias", "body.13.rdb2.conv3.weight", "body.13.rdb2.conv3.bias", "body.13.rdb2.conv4.weight", "body.13.rdb2.conv4.bias", "body.13.rdb2.conv5.weight", "body.13.rdb2.conv5.bias", "body.13.rdb3.conv1.weight", "body.13.rdb3.conv1.bias", "body.13.rdb3.conv2.weight", "body.13.rdb3.conv2.bias", "body.13.rdb3.conv3.weight", "body.13.rdb3.conv3.bias", "body.13.rdb3.conv4.weight", "body.13.rdb3.conv4.bias", "body.13.rdb3.conv5.weight", "body.13.rdb3.conv5.bias", "body.14.rdb1.conv1.weight", "body.14.rdb1.conv1.bias", "body.14.rdb1.conv2.weight", "body.14.rdb1.conv2.bias", "body.14.rdb1.conv3.weight", "body.14.rdb1.conv3.bias", "body.14.rdb1.conv4.weight", "body.14.rdb1.conv4.bias", "body.14.rdb1.conv5.weight", "body.14.rdb1.conv5.bias", "body.14.rdb2.conv1.weight", "body.14.rdb2.conv1.bias", "body.14.rdb2.conv2.weight", "body.14.rdb2.conv2.bias", "body.14.rdb2.conv3.weight", "body.14.rdb2.conv3.bias", "body.14.rdb2.conv4.weight", "body.14.rdb2.conv4.bias", "body.14.rdb2.conv5.weight", "body.14.rdb2.conv5.bias", "body.14.rdb3.conv1.weight", "body.14.rdb3.conv1.bias", "body.14.rdb3.conv2.weight", "body.14.rdb3.conv2.bias", "body.14.rdb3.conv3.weight", "body.14.rdb3.conv3.bias", "body.14.rdb3.conv4.weight", "body.14.rdb3.conv4.bias", "body.14.rdb3.conv5.weight", "body.14.rdb3.conv5.bias", "body.15.rdb1.conv1.weight", "body.15.rdb1.conv1.bias", "body.15.rdb1.conv2.weight", "body.15.rdb1.conv2.bias", "body.15.rdb1.conv3.weight", "body.15.rdb1.conv3.bias", "body.15.rdb1.conv4.weight", "body.15.rdb1.conv4.bias", "body.15.rdb1.conv5.weight", "body.15.rdb1.conv5.bias", "body.15.rdb2.conv1.weight", "body.15.rdb2.conv1.bias", "body.15.rdb2.conv2.weight", "body.15.rdb2.conv2.bias", "body.15.rdb2.conv3.weight", "body.15.rdb2.conv3.bias", "body.15.rdb2.conv4.weight", "body.15.rdb2.conv4.bias", "body.15.rdb2.conv5.weight", "body.15.rdb2.conv5.bias", "body.15.rdb3.conv1.weight", "body.15.rdb3.conv1.bias", "body.15.rdb3.conv2.weight", "body.15.rdb3.conv2.bias", "body.15.rdb3.conv3.weight", "body.15.rdb3.conv3.bias", "body.15.rdb3.conv4.weight", "body.15.rdb3.conv4.bias", "body.15.rdb3.conv5.weight", "body.15.rdb3.conv5.bias", "body.16.rdb1.conv1.weight", "body.16.rdb1.conv1.bias", "body.16.rdb1.conv2.weight", "body.16.rdb1.conv2.bias", "body.16.rdb1.conv3.weight", "body.16.rdb1.conv3.bias", "body.16.rdb1.conv4.weight", "body.16.rdb1.conv4.bias", "body.16.rdb1.conv5.weight", "body.16.rdb1.conv5.bias", "body.16.rdb2.conv1.weight", "body.16.rdb2.conv1.bias", "body.16.rdb2.conv2.weight", "body.16.rdb2.conv2.bias", "body.16.rdb2.conv3.weight", "body.16.rdb2.conv3.bias", "body.16.rdb2.conv4.weight", "body.16.rdb2.conv4.bias", "body.16.rdb2.conv5.weight", "body.16.rdb2.conv5.bias", "body.16.rdb3.conv1.weight", "body.16.rdb3.conv1.bias", "body.16.rdb3.conv2.weight", "body.16.rdb3.conv2.bias", "body.16.rdb3.conv3.weight", "body.16.rdb3.conv3.bias", "body.16.rdb3.conv4.weight", "body.16.rdb3.conv4.bias", "body.16.rdb3.conv5.weight", "body.16.rdb3.conv5.bias", "body.17.rdb1.conv1.weight", "body.17.rdb1.conv1.bias", "body.17.rdb1.conv2.weight", "body.17.rdb1.conv2.bias", "body.17.rdb1.conv3.weight", "body.17.rdb1.conv3.bias", "body.17.rdb1.conv4.weight", "body.17.rdb1.conv4.bias", "body.17.rdb1.conv5.weight", "body.17.rdb1.conv5.bias", "body.17.rdb2.conv1.weight", "body.17.rdb2.conv1.bias", "body.17.rdb2.conv2.weight", "body.17.rdb2.conv2.bias", "body.17.rdb2.conv3.weight", "body.17.rdb2.conv3.bias", "body.17.rdb2.conv4.weight", "body.17.rdb2.conv4.bias", "body.17.rdb2.conv5.weight", "body.17.rdb2.conv5.bias", "body.17.rdb3.conv1.weight", "body.17.rdb3.conv1.bias", "body.17.rdb3.conv2.weight", "body.17.rdb3.conv2.bias", "body.17.rdb3.conv3.weight", "body.17.rdb3.conv3.bias", "body.17.rdb3.conv4.weight", "body.17.rdb3.conv4.bias", "body.17.rdb3.conv5.weight", "body.17.rdb3.conv5.bias", "body.18.rdb1.conv1.weight", "body.18.rdb1.conv1.bias", "body.18.rdb1.conv2.weight", "body.18.rdb1.conv2.bias", "body.18.rdb1.conv3.weight", "body.18.rdb1.conv3.bias", "body.18.rdb1.conv4.weight", "body.18.rdb1.conv4.bias", "body.18.rdb1.conv5.weight", "body.18.rdb1.conv5.bias", "body.18.rdb2.conv1.weight", "body.18.rdb2.conv1.bias", "body.18.rdb2.conv2.weight", "body.18.rdb2.conv2.bias", "body.18.rdb2.conv3.weight", "body.18.rdb2.conv3.bias", "body.18.rdb2.conv4.weight", "body.18.rdb2.conv4.bias", "body.18.rdb2.conv5.weight", "body.18.rdb2.conv5.bias", "body.18.rdb3.conv1.weight", "body.18.rdb3.conv1.bias", "body.18.rdb3.conv2.weight", "body.18.rdb3.conv2.bias", "body.18.rdb3.conv3.weight", "body.18.rdb3.conv3.bias", "body.18.rdb3.conv4.weight", "body.18.rdb3.conv4.bias", "body.18.rdb3.conv5.weight", "body.18.rdb3.conv5.bias", "body.19.rdb1.conv1.weight", "body.19.rdb1.conv1.bias", "body.19.rdb1.conv2.weight", "body.19.rdb1.conv2.bias", "body.19.rdb1.conv3.weight", "body.19.rdb1.conv3.bias", "body.19.rdb1.conv4.weight", "body.19.rdb1.conv4.bias", "body.19.rdb1.conv5.weight", "body.19.rdb1.conv5.bias", "body.19.rdb2.conv1.weight", "body.19.rdb2.conv1.bias", "body.19.rdb2.conv2.weight", "body.19.rdb2.conv2.bias", "body.19.rdb2.conv3.weight", "body.19.rdb2.conv3.bias", "body.19.rdb2.conv4.weight", "body.19.rdb2.conv4.bias", "body.19.rdb2.conv5.weight", "body.19.rdb2.conv5.bias", "body.19.rdb3.conv1.weight", "body.19.rdb3.conv1.bias", "body.19.rdb3.conv2.weight", "body.19.rdb3.conv2.bias", "body.19.rdb3.conv3.weight", "body.19.rdb3.conv3.bias", "body.19.rdb3.conv4.weight", "body.19.rdb3.conv4.bias", "body.19.rdb3.conv5.weight", "body.19.rdb3.conv5.bias", "body.20.rdb1.conv1.weight", "body.20.rdb1.conv1.bias", "body.20.rdb1.conv2.weight", "body.20.rdb1.conv2.bias", "body.20.rdb1.conv3.weight", "body.20.rdb1.conv3.bias", "body.20.rdb1.conv4.weight", "body.20.rdb1.conv4.bias", "body.20.rdb1.conv5.weight", "body.20.rdb1.conv5.bias", "body.20.rdb2.conv1.weight", "body.20.rdb2.conv1.bias", "body.20.rdb2.conv2.weight", "body.20.rdb2.conv2.bias", "body.20.rdb2.conv3.weight", "body.20.rdb2.conv3.bias", "body.20.rdb2.conv4.weight", "body.20.rdb2.conv4.bias", "body.20.rdb2.conv5.weight", "body.20.rdb2.conv5.bias", "body.20.rdb3.conv1.weight", "body.20.rdb3.conv1.bias", "body.20.rdb3.conv2.weight", "body.20.rdb3.conv2.bias", "body.20.rdb3.conv3.weight", "body.20.rdb3.conv3.bias", "body.20.rdb3.conv4.weight", "body.20.rdb3.conv4.bias", "body.20.rdb3.conv5.weight", "body.20.rdb3.conv5.bias", "body.21.rdb1.conv1.weight", "body.21.rdb1.conv1.bias", "body.21.rdb1.conv2.weight", "body.21.rdb1.conv2.bias", "body.21.rdb1.conv3.weight", "body.21.rdb1.conv3.bias", "body.21.rdb1.conv4.weight", "body.21.rdb1.conv4.bias", "body.21.rdb1.conv5.weight", "body.21.rdb1.conv5.bias", "body.21.rdb2.conv1.weight", "body.21.rdb2.conv1.bias", "body.21.rdb2.conv2.weight", "body.21.rdb2.conv2.bias", "body.21.rdb2.conv3.weight", "body.21.rdb2.conv3.bias", "body.21.rdb2.conv4.weight", "body.21.rdb2.conv4.bias", "body.21.rdb2.conv5.weight", "body.21.rdb2.conv5.bias", "body.21.rdb3.conv1.weight", "body.21.rdb3.conv1.bias", "body.21.rdb3.conv2.weight", "body.21.rdb3.conv2.bias", "body.21.rdb3.conv3.weight", "body.21.rdb3.conv3.bias", "body.21.rdb3.conv4.weight", "body.21.rdb3.conv4.bias", "body.21.rdb3.conv5.weight", "body.21.rdb3.conv5.bias", "body.22.rdb1.conv1.weight", "body.22.rdb1.conv1.bias", "body.22.rdb1.conv2.weight", "body.22.rdb1.conv2.bias", "body.22.rdb1.conv3.weight", "body.22.rdb1.conv3.bias", "body.22.rdb1.conv4.weight", "body.22.rdb1.conv4.bias", "body.22.rdb1.conv5.weight", "body.22.rdb1.conv5.bias", "body.22.rdb2.conv1.weight", "body.22.rdb2.conv1.bias", "body.22.rdb2.conv2.weight", "body.22.rdb2.conv2.bias", "body.22.rdb2.conv3.weight", "body.22.rdb2.conv3.bias", "body.22.rdb2.conv4.weight", "body.22.rdb2.conv4.bias", "body.22.rdb2.conv5.weight", "body.22.rdb2.conv5.bias", "body.22.rdb3.conv1.weight", "body.22.rdb3.conv1.bias", "body.22.rdb3.conv2.weight", "body.22.rdb3.conv2.bias", "body.22.rdb3.conv3.weight", "body.22.rdb3.conv3.bias", "body.22.rdb3.conv4.weight", "body.22.rdb3.conv4.bias", "body.22.rdb3.conv5.weight", "body.22.rdb3.conv5.bias", "conv_body.weight", "conv_body.bias", "conv_up1.weight", "conv_up1.bias", "conv_up2.weight", "conv_up2.bias", "conv_hr.weight", "conv_hr.bias", "conv_last.weight", "conv_last.bias".
[ ]

When running the model on 384 * 384 size imaged, I get the following error from VGG discriminator in the ESRGAN

assert x.size(2) == 128 and x.size(3) == 128, (f'Input spatial size must be 128x128, '
AssertionError: Input spatial size must be 128x128, but received torch.Size([4, 3, 384, 384]).

Are these models available here:

https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG

I tried to use the ESRGAN models and getting following error:

2021-07-17 21:21:57,934 INFO: Loading VGGStyleDiscriminator128 model from /media/vikas/972ed3bd-d92e-4181-a3db-34137c27f780/vikas/work/john/wav2lip-hq-main/experiments/001_ESRGAN_x4_f64b23_custom16k_500k_B16G1_wandb/models/net_d_67500.pth.
2021-07-17 21:21:57,988 WARNING: Current net - loaded net:
2021-07-17 21:21:57,988 WARNING:   bn0_1.bias
2021-07-17 21:21:57,988 WARNING:   bn0_1.num_batches_tracked
2021-07-17 21:21:57,988 WARNING:   bn0_1.running_mean
2021-07-17 21:21:57,988 WARNING:   bn0_1.running_var
..
..
..
2021-07-17 21:21:58,014 WARNING:   conv_hr.weight
2021-07-17 21:21:58,014 WARNING:   conv_last.bias
2021-07-17 21:21:58,014 WARNING:   conv_last.weight
2021-07-17 21:21:58,014 WARNING:   conv_up1.bias
2021-07-17 21:21:58,014 WARNING:   conv_up1.weight
2021-07-17 21:21:58,014 WARNING:   conv_up2.bias
2021-07-17 21:21:58,014 WARNING:   conv_up2.weight
Traceback (most recent call last):
  File "basicsr/train.py", line 221, in <module>
    train_pipeline(root_path)
  File "basicsr/train.py", line 138, in train_pipeline
    model = build_model(opt)
  File "/media/vikas/972ed3bd-d92e-4181-a3db-34137c27f780/vikas/work/john/wav2lip-hq-main/basicsr/models/__init__.py", line 27, in build_model
    model = MODEL_REGISTRY.get(opt['model_type'])(opt)
  File "/media/vikas/972ed3bd-d92e-4181-a3db-34137c27f780/vikas/work/john/wav2lip-hq-main/basicsr/models/sr_model.py", line 32, in __init__
    self.init_training_settings()
  File "/media/vikas/972ed3bd-d92e-4181-a3db-34137c27f780/vikas/work/john/wav2lip-hq-main/basicsr/models/srgan_model.py", line 42, in init_training_settings
    self.load_network(self.net_d, load_path, self.opt['path'].get('strict_load_d', True))
  File "/media/vikas/972ed3bd-d92e-4181-a3db-34137c27f780/vikas/work/john/wav2lip-hq-main/basicsr/models/base_model.py", line 261, in load_network
    net.load_state_dict(load_net, strict=strict)
  File "/home/vikas/anaconda3/envs/py36/lib/python3.6/site-packages/torch/nn/modules/module.py", line 1407, in load_state_dict
    self.__class__.__name__, "\n\t".join(error_msgs)))

https://www.youtube.com/watch?v=Kwhqj93wyXU

I removed the line
resume_state: checkpoints/pretrained.state from train_basicsr.yml

Then got the error about 128 x 128 dimensions. So I resized "hq" images to 128x128 and "lq" images to 32x32.

Now I am getting error
LQ (32, 32) is smaller than patch size (96, 96)
in
/basicsr/data/transforms.py", line 59, in paired_random_crop

Where exactly need to set this patch size?
Has anyone got the training working without the pretrained state?

How to solve the problem of lips moving too fast when using Chinese audio samples and synthesized video

I've tested with JPG / PNG and a VIDEO and I get the same result:

• It shows parts of the face in different colors, like it's mapping the areas of nose, face, etc.. in different colors and it's not showing the actual clean final result
• It is very blurry (like the original Wav2Lip)
• The more the video progress the more blurry it gets
• As the video result progress sometimes mouth/teeth appears on random places many times = many mouths
• I've tested on short videos (few seconds 3 - 45 seconds long) I get the same results
• I used English Audio (tried different audios just in case) I get the same results
• I tried different resolutions, it doesn't matter I still get the same results

I followed all the instructions using Anaconda on Windows 10 machine, I created a dedicated 'Wav2Lip-HQ' environment, installed using the requirement file and downloaded all the needed models and placed them where needed, also did the 'wget' one, but still can't get a clean result.

Did I miss something?

Can somebody explain how can I fix that issue and make it works?
Thanks ahead!

Using cuda for inference.
Extracting raw audio...
ffmpeg version 4.2.7-0ubuntu0.1 Copyright (c) 2000-2022 the FFmpeg developers
built with gcc 9 (Ubuntu 9.4.0-1ubuntu1~20.04.1)
configuration: --prefix=/usr --extra-version=0ubuntu0.1 --toolchain=hardened --libdir=/usr/lib/x86_64-linux-gnu --incdir=/usr/include/x86_64-linux-gnu --arch=amd64 --enable-gpl --disable-stripping --enable-avresample --disable-filter=resample --enable-avisynth --enable-gnutls --enable-ladspa --enable-libaom --enable-libass --enable-libbluray --enable-libbs2b --enable-libcaca --enable-libcdio --enable-libcodec2 --enable-libflite --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libgme --enable-libgsm --enable-libjack --enable-libmp3lame --enable-libmysofa --enable-libopenjpeg --enable-libopenmpt --enable-libopus --enable-libpulse --enable-librsvg --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libssh --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwavpack --enable-libwebp --enable-libx265 --enable-libxml2 --enable-libxvid --enable-libzmq --enable-libzvbi --enable-lv2 --enable-omx --enable-openal --enable-opencl --enable-opengl --enable-sdl2 --enable-libdc1394 --enable-libdrm --enable-libiec61883 --enable-nvenc --enable-chromaprint --enable-frei0r --enable-libx264 --enable-shared
libavutil 56. 31.100 / 56. 31.100
libavcodec 58. 54.100 / 58. 54.100
libavformat 58. 29.100 / 58. 29.100
libavdevice 58. 8.100 / 58. 8.100
libavfilter 7. 57.100 / 7. 57.100
libavresample 4. 0. 0 / 4. 0. 0
libswscale 5. 5.100 / 5. 5.100
libswresample 3. 5.100 / 3. 5.100
libpostproc 55. 5.100 / 55. 5.100
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'videos/shulman_30s.mp4':
Metadata:
major_brand : isom
minor_version : 512
compatible_brands: isomiso2avc1mp41
encoder : Lavf57.83.100
Duration: 00:00:30.88, start: 0.000000, bitrate: 165 kb/s
Stream #0:0(und): Video: h264 (Main) (avc1 / 0x31637661), yuv420p(tv, bt709), 256x144 [SAR 1:1 DAR 16:9], 30 kb/s, 25 fps, 25 tbr, 12800 tbn, 50 tbc (default)
Metadata:
handler_name : VideoHandler
Stream #0:1(eng): Audio: aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 127 kb/s (default)
Metadata:
handler_name : SoundHandler
Stream mapping:
Stream #0:1 -> #0:0 (aac (native) -> pcm_s16le (native))
Press [q] to stop, [?] for help
Output #0, wav, to 'temp/temp.wav':
Metadata:
major_brand : isom
minor_version : 512
compatible_brands: isomiso2avc1mp41
ISFT : Lavf58.29.100
Stream #0:0(eng): Audio: pcm_s16le ([1][0][0][0] / 0x0001), 44100 Hz, stereo, s16, 1411 kb/s (default)
Metadata:
handler_name : SoundHandler
encoder : Lavc58.54.100 pcm_s16le
size= 5168kB time=00:00:30.00 bitrate=1411.2kbits/s speed= 792x
video:0kB audio:5168kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.001474%
(80, 2401)
Length of mel chunks: 895
0% 0/7 [00:00<?, ?it/s]Reading video frames from start...
Loading segmentation network...
0% 0/7 [01:15<?, ?it/s]
Traceback (most recent call last):
File "inference.py", line 355, in
main()
File "inference.py", line 296, in main
seg_net = init_parser(args.segmentation_path)
File "/content/wav2lip-hq/face_parsing/swap.py", line 13, in init_parser
net.load_state_dict(torch.load(pth_path))
File "/usr/local/lib/python3.8/dist-packages/torch/serialization.py", line 795, in load
return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
File "/usr/local/lib/python3.8/dist-packages/torch/serialization.py", line 1002, in _legacy_load
magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, '<'.

Hi!
Thank you so much for making work public : )

I have one question here. In the original GitHub repository, there are two wav2lip models provided-

1. wav2lip.pth
2. wav2lip_gan.pth

But in your repository, I could not find the link to the model without GAN. The model without GAN, as mentioned in the original paper, claims to perform better lip sync, but has lower quality.

It would be great it you could make the wav2lip.pth (pytorch model) public too.

Thanks again : )