Missing Files and Incompatibilities

[csimlinger]

We would love to play around with your tool but we are encountering multiple issues while trying to install and set up the Lip_Wise repository, even after following the provided instructions. Below is a summary of the steps we followed and the issues encountered:

Steps Followed:

1. Cloned the repository:

git clone https://github.com/pawansharmaaaa/Lip_Wise

2. Created a new Conda environment with Python 3.10:

conda create -n lipwise_env python=3.10
conda activate lipwise_env

3. Made the setup.sh and launch.sh executable:

chmod +x ./setup.sh
chmod +x ./launch.sh

4. Attempted to install dependencies via setup.sh:
   There were issues with torchvision==0.16.2, so we downgraded to torchvision==0.16.0 and adjusted the versions in requirements.txt to match the required dependencies.

5. Installed specific package versions based on the minimum requirements to avoid incompatibility issues.

Current Issues:

1. Missing torchvision.transforms.functional_tensor:
   After running the script, we encountered the following error:

ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'
We tried using several versions of torchvision (0.16.0, 0.16.2), but this issue persists.

2. basicsr.test Not Found:
   Another error we encountered is:

ModuleNotFoundError: No module named 'basicsr.test'
We installed basicsr==1.4.2, but the error remains. It seems the required module or file is missing from the installation.

3. cp: cannot stat 'archs/*': No such file or directory:
   During the setup process, the script tries to copy files from an archs/ directory, which does not exist in the repository:

cp: cannot stat 'archs/*': No such file or directory

4. Dependency conflicts with numpy and gradio:
   We faced dependency conflicts related to gradio and numpy:

• gradio requires numpy~=1.0, but other dependencies like torch require numpy>=1.26.2.

• numba requires numpy<1.27, causing version conflicts.

Questions:

1. How can we resolve the missing torchvision.transforms.functional_tensor issue? Is there a specific version of torchvision that we should use?
2. What is the correct version of basicsr to resolve the basicsr.test issue?
3. Is the missing archs/* directory an oversight, or is there another path to resolve this?
4. How do we manage the dependency conflicts, especially between numpy, torch, gradio, and numba?
5. We would greatly appreciate any guidance or advice on how to resolve these issues and get the project running.

Thanks a lot!