This message was deleted.

I add a simple model code that reproduces the issue.

import torch
from torch import nn
from transformers import Wav2Vec2Model


class EmbeddedModel(nn.Module):
    def __init__(self):
        super(EmbeddedModel, self).__init__()

        self.model = Wav2Vec2Model.from_pretrained("facebook/wav2vec2-base-960h")

    def forward(self, x):
        return self.model(x)

Hey! Thanks for the findings, we will take a look into it