Skip to content

Enabling CPU runs causes train.py to fail if bash scripts/train_tacotron2.sh or / train_waveglow.sh is called  #496

Description

@sprenkamp

Related to Model/Framework(s)
(e.g. GNMT/PyTorch or FasterTransformer/All)

Describe the bug
Since the feature -- cpu-run was implemeted train.py breaks for me, see the error below:
File "train.py", line 501, in
main()
File "train.py", line 335, in main
uniform_initialize_bn_weight=not args.disable_uniform_initialize_bn_weight)
TypeError: get_model() got an unexpected keyword argument 'to_cuda'

I guess this due to the changes in get_model() to enable CPU runs and it should be easily fixed with changing the input parameters of get_model()

To Reproduce
Steps to reproduce the behavior:

I followed the quick setup guide within the repository.

Expected behavior

get_model is able to to load model for training.

Environment
Please provide at least:

  • Container version: torch: 1.5.0a0+8f84ded
  • GPUs in the system: 4xTesla T4
  • CUDA driver version: release 10.2, V10.2.89

Comment
First time I am creating an issue. Sorry if something is not clear.

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions