ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'

[vter00]

ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'

[DFMlaozhu]

I encountered the same problem as you, have you solved it?

[andradeofc]

I have the same problem

[arielweinberger]

Go to file /usr/local/lib/python3.10/dist-packages/basicsr/data/degradations.py and change line number 8 to:

from torchvision.transforms.functional import rgb_to_grayscale

Got the solution from AUTOMATIC1111/stable-diffusion-webui#13985

[netxor17]

Try to uninstall and then install the latest version of torchvision. (0.17.1)
pip install torchvision

[Matrix-X]

which version of torch, torchvision, torchaudio is recommended and run success on Mac M1 ?

[Matrix-X]

Try to uninstall and then install the latest version of torchvision. (0.17.1) pip install torchvision

❯ pip list | grep torch
torch 2.2.1
torchaudio 2.2.1
torchvision 0.17.1
❯ python inference.py --face examples/face/1.mp4 --audio examples/audio/1.wav --outfile results/1_1.mp4
Traceback (most recent call last):
File "inference.py", line 16, in
from third_part.GPEN.gpen_face_enhancer import FaceEnhancement
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/third_part/GPEN/gpen_face_enhancer.py", line 11, in
from utils.inference_utils import Laplacian_Pyramid_Blending_with_mask
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/utils/inference_utils.py", line 5, in
from models import load_network, load_DNet
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/init.py", line 2, in
from models.DNet import DNet
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/DNet.py", line 10, in
from models.base_blocks import LayerNorm2d, ADAINHourglass, FineEncoder, FineDecoder
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/base_blocks.py", line 9, in
from basicsr.archs.arch_util import default_init_weights
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/basicsr/init.py", line 4, in
from .data import *
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/basicsr/data/init.py", line 22, in
_dataset_modules = [importlib.import_module(f'basicsr.data.{file_name}') for file_name in dataset_filenames]
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/basicsr/data/init.py", line 22, in
_dataset_modules = [importlib.import_module(f'basicsr.data.{file_name}') for file_name in dataset_filenames]
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/importlib/init.py", line 127, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/basicsr/data/realesrgan_dataset.py", line 11, in
from basicsr.data.degradations import circular_lowpass_kernel, random_mixed_kernels
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/basicsr/data/degradations.py", line 8, in
from torchvision.transforms.functional_tensor import rgb_to_grayscale
ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'

[Matrix-X]

which version of torch, torchvision, torchaudio is recommended and run success on Mac M1 ?

pip install torch==1.9.0 torchvision==0.10.0 torchaudio==0.9.0

❯ python inference.py --face examples/face/1.mp4 --audio examples/audio/1.wav --outfile results/1_1.mp4
[Info] Using cpu for inference.
[Step 0] Number of frames available for inference: 135
[Step 1] Using saved landmarks.
[Step 2] 3DMM Extraction In Video:: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 135/135 [00:05<00:00, 26.32it/s]
using expression center
Load checkpoint from: checkpoints/DNet.pt
Load checkpoint from: checkpoints/LNet.pth
Load checkpoint from: checkpoints/ENet.pth
[Step 3] Using saved stabilized video.
[Step 4] Load audio; Length of mel chunks: 109
[Step 5] Reference Enhancement: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 109/109 [08:27<00:00, 4.65s/it]
landmark Det:: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 109/109 [00:46<00:00, 2.32it/s]
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 109/109 [00:00<00:00, 41943.04it/s]
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 109/109 [00:00<00:00, 1026.23it/s]
FaceDet:: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 28/28 [01:11<00:00, 2.55s/it]
[Step 6] Lip Synthesis:: 0%| | 0/7 [02:04<?, ?it/s]
Traceback (most recent call last):
File "inference.py", line 345, in
main()
File "inference.py", line 221, in main
pred, low_res = model(mel_batch, img_batch, reference)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/ENet.py", line 113, in forward
low_res_img = self.low_res(audio_sequences, LNet_input)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/LNet.py", line 132, in forward
_outputs = self.decoder(vis_feat, audio_feat)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/LNet.py", line 73, in forward
out = res_model(out, z)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/base_blocks.py", line 425, in forward
x = model(x, z)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/base_blocks.py", line 404, in forward
x_l, x_g = self.conv1((x_l, x_g), z)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/base_blocks.py", line 383, in forward
x_l, x_g = self.ffc(x)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/ffc.py", line 231, in forward
out_xg = self.convl2g(x_l) * l2g_gate + self.convg2g(x_g)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/ffc.py", line 157, in forward
output = self.fu(x)
File "/opt/anaconda3/envs/video_retalking/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/Volumes/AISpace/Workspace/DigitalHuman/video-retalking/models/ffc.py", line 99, in forward
ffted = torch.fft.rfftn(x, dim=fft_dim, norm=self.fft_norm)
RuntimeError: fft: ATen not compiled with MKL support

[tomy128]

This is an bug for package basicsr==1.4.2, see this for details: https://github.com/XPixelGroup/BasicSR/pull/650/files

[vettorazi]

wow, that was super sketchy... but changing the 2D thing in both files + changing how the degradation file imports tourchvision (torchvision.transforms.functional import rgb_to_grayscale) + changing the requirements.txt
worked for me! basically, what I'm saying is: if you try the first and second solves and nothing worked.. keep trying! some hack will fix this thing haha