Use reshape instead of view

[vincentqb]

Following #409, view is used instead of reshape. When running the tutorial, the following error occurs as mentioned here:

Apr 22 18:46:35 Traceback (most recent call last):
Apr 22 18:46:35   File "/opt/conda/lib/python3.6/site-packages/sphinx_gallery/gen_rst.py", line 394, in _memory_usage
Apr 22 18:46:35     out = func()
Apr 22 18:46:35   File "/opt/conda/lib/python3.6/site-packages/sphinx_gallery/gen_rst.py", line 382, in __call__
Apr 22 18:46:35     exec(self.code, self.globals)
Apr 22 18:46:35   File "/var/lib/jenkins/workspace/beginner_source/audio_preprocessing_tutorial.py", line 216, in <module>
Apr 22 18:46:35     computed = torchaudio.functional.compute_deltas(specgram, win_length=3)
Apr 22 18:46:35   File "/opt/conda/lib/python3.6/site-packages/torchaudio/functional.py", line 1304, in compute_deltas
Apr 22 18:46:35     specgram = specgram.view(1, -1, shape[-1])
Apr 22 18:46:35 RuntimeError: view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.

Workaround:

• add specgram = specgram.contiguous(), see make tensor contiguous tutorials#963

Proposed solution:

• revert replace reshape by view #409

[vincentqb]

cc @cpuhrsch for comment

[vincentqb]

Closed by #594