output_chanels and loading

[NikolaAndro]

I was loading xception-full model and got an error:

taskaugment_venv/lib/python3.10/site-packages/torch/nn/init.py:405: UserWarning: Initializing zero-element tensors is a no-op
  warnings.warn("Initializing zero-element tensors is a no-op")
Traceback (most recent call last):
  File "taskaugment/MTVulnerability/experiments/attacks/adv_attack.py", line 110, in <module>
    run(args)
  File "taskaugment/MTVulnerability/experiments/attacks/adv_attack.py", line 58, in run
    model = load_model(args)
  File "taskaugment/MTVulnerability/experiments/attacks/__init__.py", line 77, in load_model
    model = XceptionTaskonomy(tasks=args.test_task_set)
  File "taskaugment/MTVulnerability/utils/xception_taskonomy_new.py", line 359, in __init__
    m.weight.data.normal_(0, math.sqrt(2. / n))
ZeroDivisionError: float division by zero

The error was an inconsistency in the variable chanels. In some places was spelled channels. I made a small change so the variable is consistently spelled incorrectly :) Then, it was able to load.

Also, when loading state_dict, it did not work without the strict parameter being set to True. So, I used your comment.

Thanks!