[bug]: ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor' torchvision 0.17 promblem #5108

KEDI103 · 2023-11-16T11:11:21Z

Is there an existing issue for this?

I have searched the existing issues

OS

Linux

GPU

amd

VRAM

16

What version did you experience this issue on?

all version contain

What happened?


Generate images with a browser-based interface
2023-11-16 00:24:48.166525: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
[2023-11-16 00:24:52,684]::[InvokeAI]::INFO --> Loaded 0 modules from /media/b_cansin/ai/ai/InvokeAI-Installer/nodes
>> patchmatch.patch_match: INFO - Compiling and loading c extensions from "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/patchmatch".
>> patchmatch.patch_match: ERROR - patchmatch failed to load or compile (Command 'make clean && make' returned non-zero exit status 2.).
>> patchmatch.patch_match: INFO - Refer to https://invoke-ai.github.io/InvokeAI/installation/060_INSTALL_PATCHMATCH/ for installation instructions.
[2023-11-16 00:24:58,694]::[InvokeAI]::INFO --> Patchmatch not loaded (nonfatal)
Traceback (most recent call last):
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/bin/invokeai-web", line 5, in <module>
    from invokeai.app.api_app import invoke_api
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/invokeai/app/api_app.py", line 40, in <module>
    from .api.dependencies import ApiDependencies
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/invokeai/app/api/dependencies.py", line 10, in <module>
    from ..services.board_images.board_images_default import BoardImagesService
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/invokeai/app/services/board_images/board_images_default.py", line 3, in <module>
    from invokeai.app.services.invoker import Invoker
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/invokeai/app/services/invoker.py", line 7, in <module>
    from .shared.graph import Graph, GraphExecutionState
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/invokeai/app/services/shared/graph.py", line 12, in <module>
    from invokeai.app.invocations import *  # noqa: F401 F403
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/invokeai/app/invocations/upscale.py", line 8, in <module>
    from basicsr.archs.rrdbnet_arch import RRDBNet
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/basicsr/__init__.py", line 4, in <module>
    from .data import *
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/basicsr/data/__init__.py", line 22, in <module>
    _dataset_modules = [importlib.import_module(f'basicsr.data.{file_name}') for file_name in dataset_filenames]
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/basicsr/data/__init__.py", line 22, in <listcomp>
    _dataset_modules = [importlib.import_module(f'basicsr.data.{file_name}') for file_name in dataset_filenames]
  File "/usr/lib/python3.10/importlib/__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/basicsr/data/realesrgan_dataset.py", line 11, in <module>
    from basicsr.data.degradations import circular_lowpass_kernel, random_mixed_kernels
  File "/media/b_cansin/ai/ai/InvokeAI-Installer/.venv/lib/python3.10/site-packages/basicsr/data/degradations.py", line 8, in <module>
    from torchvision.transforms.functional_tensor import rgb_to_grayscale
ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'

Screenshots

No response

Additional context

The cure is:
AUTOMATIC1111/stable-diffusion-webui#13985 (comment)
I tested for AUTOMATIC1111 its working but for invokeai it hide venv folder and only accesable by dev console could be need bit help to edit it.
For real fix awaiting approval:
XPixelGroup/BasicSR#650

Contact Details

discord: b_cansin

The text was updated successfully, but these errors were encountered:

psychedelicious · 2023-11-16T21:22:36Z

we've pinned torchvision to 0.16.0 which does not have this problem. will be in the next release, should fix this for you. hopefully the upstream PR gets merged.

KEDI103 · 2023-11-16T21:40:40Z

we've pinned torchvision to 0.16.0 which does not have this problem. will be in the next release, should fix this for you. hopefully the upstream PR gets merged.

I edited myself it fixed doing this
AUTOMATIC1111/stable-diffusion-webui#13985 (comment)
But I want to inform here too. I need torch 2.2 with torchvision 0.17 because of this
pytorch/pytorch#103973 (comment)

psychedelicious · 2023-11-16T21:46:07Z

@KEDI103 torch 2.2 doesn't exist. torch 2.1.0 is fine with torchvision 0.16.0.

While the fix you applied works, it's not a good way to fix this issue. If we end up needing torchvision 0.17.0, we can figure out a solution (possibly just a fork of basicsr with the fix)

KEDI103 · 2023-11-16T21:57:47Z

@KEDI103 torch 2.2 doesn't exist. torch 2.1.0 is fine with torchvision 0.16.0.

While the fix you applied works, it's not a good way to fix this issue. If we end up needing torchvision 0.17.0, we can figure out a solution (possibly just a fork of basicsr with the fix)

heres torch 2.2 also works with invokeai too

psychedelicious · 2023-11-16T22:01:18Z

the latest torch release is 2.1.1, just released yesterday

you appear to have a dev/nightly build of torch installed. invokeai doesn't use those.

KEDI103 · 2023-11-16T22:05:29Z

the latest torch release is 2.1.1, just released yesterday

you appear to have a dev/nightly build of torch installed. invokeai doesn't use those.

I have force to use it we got for months of test to deal with pci atomics problems for ROCM
pytorch/pytorch#103973 (comment)

Also can we got options in updater won't always force uninstall torch because it give me additional steps to delete and reinstall my forced torch versions.

psychedelicious · 2023-11-17T02:01:54Z

What kind of options do you need? Do you mean to install different versions of certain packages?

KEDI103 · 2023-11-17T02:15:15Z

What kind of options do you need? Do you mean to install different versions of certain packages?

I meant when I update invoke update detect my pytorch version and uninstall it everytime and reinstall the broken one for me. And I need to reinstall my needed pytorch version from dev console.

If there is option to keep the installed pytorch would be so awesome. It saves so much time for me.

psychedelicious · 2023-11-17T02:32:29Z

@lstein
There's an issue with rocm that is resolved on a dev release of PyTorch. @KEDI103 (and other users with certain gpus) will need to use 2.2dev (or whatever other build) along with torchvision 0.17.0.

Can we provide a special optional package group for this? Say, ".[rocm-fix]", something like that? I'm not sure how this stuff all interacts.

@KEDI103 which gpus are affected?

fxzjshm · 2023-11-17T15:12:29Z

Affected GPUs are gfx906 based GPUs with no PCIe atomics, e.g. Radeon VII, Instinct MI50, etc.

KEDI103 · 2023-11-17T16:02:08Z

Affected GPUs are gfx906 based GPUs with no PCIe atomics, e.g. Radeon VII, Instinct MI50, etc.

Actually this not gpu problem its pci atomics missing problem.

KEDI103 · 2023-11-17T20:23:32Z

@lstein There's an issue with rocm that is resolved on a dev release of PyTorch. @KEDI103 (and other users with certain gpus) will need to use 2.2dev (or whatever other build) along with torchvision 0.17.0.

Can we provide a special optional package group for this? Say, ".[rocm-fix]", something like that? I'm not sure how this stuff all interacts.

@KEDI103 which gpus are affected?

Next week nighty pre build will show us. Also It not gpu problem its motherboard+cpu missing pci atomics support.
pytorch/pytorch#103973 (comment)

psychedelicious · 2023-11-17T20:58:31Z

Ah, I see. Thanks for clarifying.

Do we have an ETA on torch 2.2.0?

KEDI103 · 2023-11-17T21:14:07Z

Ah, I see. Thanks for clarifying.

Do we have an ETA on torch 2.2.0?

well right now I gen at torch-2.2.0.dev20231114+rocm5.7-cp310-cp310-linux_x86_64 with torchvision-0.17.0+rocm5.7-cp310-cp310-linux_x86_64
If this going to merge with pre I think yes

psychedelicious · 2023-11-17T22:39:58Z

Sorry, what I meant is, do we know when torch 2.2.0 is expected to be released?

KEDI103 · 2023-11-17T22:50:18Z

Sorry, what I meant is, do we know when torch 2.2.0 is expected to be released?

Maybe next week because what I am using wheel gona merge with offical nighty pre build.

psychedelicious · 2023-11-17T23:03:21Z

Usually there are only a few torch minor version releases per year (they say 3 per year is typical). We just had 2.1.0 at the beginning of october, so I'd be surprised if we get 2.2.0 so soon after.

If we can just wait until torch 2.2.0 is released, that seems a lot simpler than updating the installation process with a new UI to install certain versions of things.

`basicsr` has a hard dependency on torchvision <= 0.16 and is unmaintained. Extract the code we need from it and remove the dep. Closes #5108

ohmerhe · 2024-04-28T03:25:03Z

Still have this issue in 2.2.2 on mac

Name: torch
Version: 2.2.2
Summary: Tensors and Dynamic neural networks in Python with strong GPU acceleration
Home-page: https://pytorch.org/
Author: PyTorch Team
Author-email: packages@pytorch.org
License: BSD-3
Location: /Users/ohmer/miniconda3/envs/MoneyPrinterTurbo/lib/python3.10/site-packages
Requires: filelock, fsspec, jinja2, networkx, sympy, typing-extensions
Required-by: basicsr, facexlib, gfpgan, realesrgan, torchaudio, torchvision

Name: torchvision
Version: 0.17.2
Summary: image and video datasets and models for torch deep learning
Home-page: https://github.com/pytorch/vision
Author: PyTorch Core Team
Author-email: soumith@pytorch.org
License: BSD
Location: /Users/ohmer/miniconda3/envs/MoneyPrinterTurbo/lib/python3.10/site-packages
Requires: numpy, pillow, torch
Required-by: basicsr, facexlib, gfpgan, realesrgan

psychedelicious · 2024-04-28T03:44:37Z

@ohmerhe You are on an old version. Please follow the instructions here to update: https://invoke-ai.github.io/InvokeAI/installation/010_INSTALL_AUTOMATED/

KEDI103 added the bug Something isn't working label Nov 16, 2023

onlyreportingissues mentioned this issue Dec 21, 2023

[Fix for] ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor' torchvision 0.17 problem vladmandic/automatic#2643

Closed

1 task

psychedelicious added a commit that referenced this issue Feb 10, 2024

feat(backend): remove dependency on basicsr

eb802d4

`basicsr` has a hard dependency on torchvision <= 0.16 and is unmaintained. Extract the code we need from it and remove the dep. Closes #5108

psychedelicious added a commit that referenced this issue Feb 10, 2024

feat(backend): remove dependency on basicsr

60aad95

`basicsr` has a hard dependency on torchvision <= 0.16 and is unmaintained. Extract the code we need from it and remove the dep. Closes #5108

psychedelicious mentioned this issue Feb 10, 2024

feat(backend): remove dependency on basicsr #5690

Merged

10 tasks

psychedelicious closed this as completed in #5690 Feb 10, 2024

psychedelicious added a commit that referenced this issue Feb 10, 2024

feat(backend): remove dependency on basicsr

c5f069a

`basicsr` has a hard dependency on torchvision <= 0.16 and is unmaintained. Extract the code we need from it and remove the dep. Closes #5108

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[bug]: ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor' torchvision 0.17 promblem #5108

[bug]: ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor' torchvision 0.17 promblem #5108

KEDI103 commented Nov 16, 2023

psychedelicious commented Nov 16, 2023

KEDI103 commented Nov 16, 2023

psychedelicious commented Nov 16, 2023 •

edited

Loading

KEDI103 commented Nov 16, 2023

psychedelicious commented Nov 16, 2023 •

edited

Loading

KEDI103 commented Nov 16, 2023

psychedelicious commented Nov 17, 2023

KEDI103 commented Nov 17, 2023

psychedelicious commented Nov 17, 2023

fxzjshm commented Nov 17, 2023 •

edited

Loading

KEDI103 commented Nov 17, 2023

KEDI103 commented Nov 17, 2023

psychedelicious commented Nov 17, 2023

KEDI103 commented Nov 17, 2023

psychedelicious commented Nov 17, 2023

KEDI103 commented Nov 17, 2023

psychedelicious commented Nov 17, 2023

ohmerhe commented Apr 28, 2024

psychedelicious commented Apr 28, 2024

[bug]: ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor' torchvision 0.17 promblem #5108

[bug]: ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor' torchvision 0.17 promblem #5108

Comments

KEDI103 commented Nov 16, 2023

Is there an existing issue for this?

OS

GPU

VRAM

What version did you experience this issue on?

What happened?

Screenshots

Additional context

Contact Details

psychedelicious commented Nov 16, 2023

KEDI103 commented Nov 16, 2023

psychedelicious commented Nov 16, 2023 • edited Loading

KEDI103 commented Nov 16, 2023

psychedelicious commented Nov 16, 2023 • edited Loading

KEDI103 commented Nov 16, 2023

psychedelicious commented Nov 17, 2023

KEDI103 commented Nov 17, 2023

psychedelicious commented Nov 17, 2023

fxzjshm commented Nov 17, 2023 • edited Loading

KEDI103 commented Nov 17, 2023

KEDI103 commented Nov 17, 2023

psychedelicious commented Nov 17, 2023

KEDI103 commented Nov 17, 2023

psychedelicious commented Nov 17, 2023

KEDI103 commented Nov 17, 2023

psychedelicious commented Nov 17, 2023

ohmerhe commented Apr 28, 2024

psychedelicious commented Apr 28, 2024

psychedelicious commented Nov 16, 2023 •

edited

Loading

psychedelicious commented Nov 16, 2023 •

edited

Loading

fxzjshm commented Nov 17, 2023 •

edited

Loading