Fix mismatch between model num_output_labels and linear weights dimension in checkpoint

While trying out your model, I stumbled upon the error message:

RuntimeError: Error(s) in loading state_dict for LaiaCRNN:
  size mismatch for linear.weight: copying a param with shape torch.Size([117, 512]) from checkpoint, the shape in current model is torch.Size([116, 512]).
  size mismatch for linear.bias: copying a param with shape torch.Size([117]) from checkpoint, the shape in current model is torch.Size([116]).

By aligning the num_output_labels to match the 117 from the checkpoint, I could use it successfully.

Thanks for the upload !