Skip to content

Training voice fails with PyTorch 2.6 #225

Description

@Leviani

By now I tried several different approaches.
Since I have a Nvidia Jetson Nano, I thought that might be a good place to run my training of a new custom voice, but it turned out to be the worst nightmare ever.
With the help of some AI tools I spend about 2 days trying to find the correct wheel and packages for Jetpack 6.2 and eventually gave up, because I never got it to run.

Next was trying to run it inside Ubuntu in WSL on my Windows 11, which took a lot of extra steps, compared to the official Readme.
Since I didn't really understand what that sentence meant "If you are running from the repo, you will need to do a dev build:" I tried ./build_monotonic_align.sh first and got to this problem #14

Then I hit another error telling me that I had the wrong cmake version and had to run

sudo apt purge --auto-remove cmake
sudo snap install cmake --classic

After running python3 setup.py build_ext --inplace I finally got the command to start, but now I am stuck with

Weights only load failed. In PyTorch 2.6, we changed the default value of the `weights_only` argument in `torch.load` from `False` to `True`. Re-running `torch.load` with `weights_only` set to `False` will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source.
Please file an issue with the following so that we can make `weights_only=True` compatible with your use case: WeightsUnpickler error:

Unsupported operand 60

Is this now an issue where you need to install very specific Versions of different Libraries to make this work ?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions