Fix pad mask traceable

[iamanigeeit]

[skip ci] xs.shape[length_dim] sometimes returns tensor instead of int, causing different device error at torch.clamp(lengths, max=max_length). See (unresolved) issue at https://discuss.pytorch.org/t/x-shape-or-x-size-return-torch-size-with-elements-of-type-tensor/153136

This commit fixes it.

[codecov]

Codecov Report

Attention: Patch coverage is 66.66667% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 76.12%. Comparing base (0854d6f) to head (449b455).

Files	Patch %	Lines
espnet/nets/pytorch_backend/nets_utils.py	66.66%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5680      +/-   ##
==========================================
+ Coverage   70.27%   76.12%   +5.84%     
==========================================
  Files         746      747       +1     
  Lines       69369    69411      +42     
==========================================
+ Hits        48752    52836    +4084     
+ Misses      20617    16575    -4042     

Flag	Coverage Δ
test_configuration_espnet2	∅ <ø> (∅)
test_integration_espnet1	62.91% <0.00%> (-0.01%)	⬇️
test_integration_espnet2	48.92% <33.33%> (?)
test_python_espnet1	18.32% <66.66%> (+<0.01%)	⬆️
test_python_espnet2	52.69% <33.33%> (+0.64%)	⬆️
test_utils	22.15% <0.00%> (+0.24%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.