Warp-ctc request for cuda_ver in CPU standalone.

[Fhrozen]

@jnishi from the update (@ysk24ok) to warp-ctc, it request for cuda version even when there is a CPU-only pc.

The output is:

Torch was not built with CUDA support, not building warp-ctc GPU extensions.
Traceback (most recent call last):
  File "setup.py", line 66, in <module>
    get_torch_version(), get_cuda_version())
  File "setup.py", line 55, in get_cuda_version
    proc = Popen(['nvcc', '--version'], stdout=PIPE, stderr=PIPE)
  File "/home/nelson/docstrings/tools/venv/lib/python3.7/subprocess.py", line 775, in __init__
    restore_signals, start_new_session)
  File "/home/nelson/docstrings/tools/venv/lib/python3.7/subprocess.py", line 1522, in _execute_child
    raise child_exception_type(errno_num, err_msg, err_filename)
FileNotFoundError: [Errno 2] No such file or directory: 'nvcc': 'nvcc'

warp-ctc/pytorch_binding/setup.py

Lines 65 to 66 in 4209ce4

	package_name = 'warpctc_pytorch{}_cuda{}'.format(
	get_torch_version(), get_cuda_version())

It will be better to set to:

if enable_gpu: 
    tag = 'cuda{}'.format(get_cuda_version())
else:
    tag = 'cpu'
package_name = 'warpctc_pytorch{}_{}'.format(
    get_torch_version(), tag)

or similar

[ysk24ok]

Thank you for the feedback.
I'll deal with this by creating warpctc_pytorch100_cpu wheel in CI.

[Fhrozen]

I just tested without any problem, thanks for the support.