Fix [PyT/DLRM] bug of model.py - #517

python3 -m dlrm.scripts.main --mode train --max_steps 500 --benchmark_warmup_steps 250 --synthetic_dataset --print_freq 1

(...snip)

Traceback (most recent call last):
  File "/usr/lib64/python3.6/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/usr/lib64/python3.6/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/root/DeepLearningExamples/PyTorch/Recommendation/DLRM/dlrm/scripts/main.py", line 509, in <module>
    app.run(main)
  File "/usr/local/lib/python3.6/site-packages/absl/app.py", line 299, in run
    _run_main(main, args)
  File "/usr/local/lib/python3.6/site-packages/absl/app.py", line 250, in _run_main
    sys.exit(main(argv))
  File "/root/DeepLearningExamples/PyTorch/Recommendation/DLRM/dlrm/scripts/main.py", line 293, in main
    train(model, loss_fn, optimizer, data_loader_train, data_loader_test, scaled_lr)
  File "/root/DeepLearningExamples/PyTorch/Recommendation/DLRM/dlrm/scripts/main.py", line 355, in train
    output = model(numerical_features, categorical_features).squeeze().float()
  File "/usr/local/lib64/python3.6/site-packages/torch/nn/modules/module.py", line 532, in __call__
    result = self.forward(*input, **kwargs)
  File "/root/DeepLearningExamples/PyTorch/Recommendation/DLRM/dlrm/model.py", line 220, in forward
    interaction_output = self._interaction(bottom_mlp_output, embedding_outputs, batch_size)
  File "/root/DeepLearningExamples/PyTorch/Recommendation/DLRM/dlrm/model.py", line 129, in _interaction
    (bottom_mlp_output, interaction_flat, interaction_padding), dim=1)
RuntimeError: Expected object of backend CUDA but got backend CPU for sequence element 2 in sequence argument at position #1 'tensors'