Text-to-speech with TorchAudio can not compile waveglow part

[malfet]

As:

torch.hub.load('NVIDIA/DeepLearningExamples:torchhub', 'nvidia_waveglow', model_math='fp32')

now returns HTTP/404:

Oct 07 00:05:10     waveglow = torch.hub.load('NVIDIA/DeepLearningExamples:torchhub', 'nvidia_waveglow', model_math='fp32')
Oct 07 00:05:10   File "/opt/conda/lib/python3.7/site-packages/torch/hub.py", line 540, in load
Oct 07 00:05:10     model = _load_local(repo_or_dir, model, *args, **kwargs)
Oct 07 00:05:10   File "/opt/conda/lib/python3.7/site-packages/torch/hub.py", line 569, in _load_local
Oct 07 00:05:10     model = entry(*args, **kwargs)
Oct 07 00:05:10   File "/var/lib/jenkins/.cache/torch/hub/NVIDIA_DeepLearningExamples_torchhub/PyTorch/SpeechSynthesis/Tacotron2/waveglow/entrypoints.py", line 95, in nvidia_waveglow
Oct 07 00:05:10     ckpt_file = _download_checkpoint(checkpoint, force_reload)
Oct 07 00:05:10   File "/var/lib/jenkins/.cache/torch/hub/NVIDIA_DeepLearningExamples_torchhub/PyTorch/SpeechSynthesis/Tacotron2/waveglow/entrypoints.py", line 73, in _download_checkpoint
Oct 07 00:05:10     urllib.request.urlretrieve(checkpoint, ckpt_file)
Oct 07 00:05:10   File "/opt/conda/lib/python3.7/urllib/request.py", line 247, in urlretrieve
Oct 07 00:05:10     with contextlib.closing(urlopen(url, data)) as fp:
Oct 07 00:05:10   File "/opt/conda/lib/python3.7/urllib/request.py", line 222, in urlopen
Oct 07 00:05:10     return opener.open(url, data, timeout)
Oct 07 00:05:10   File "/opt/conda/lib/python3.7/urllib/request.py", line 531, in open
Oct 07 00:05:10     response = meth(req, response)
Oct 07 00:05:10   File "/opt/conda/lib/python3.7/urllib/request.py", line 641, in http_response
Oct 07 00:05:10     'http', request, response, code, msg, hdrs)
Oct 07 00:05:10   File "/opt/conda/lib/python3.7/urllib/request.py", line 569, in error
Oct 07 00:05:10     return self._call_chain(*args)
Oct 07 00:05:10   File "/opt/conda/lib/python3.7/urllib/request.py", line 503, in _call_chain
Oct 07 00:05:10     result = func(*args)
Oct 07 00:05:10   File "/opt/conda/lib/python3.7/urllib/request.py", line 649, in http_error_default
Oct 07 00:05:10     raise HTTPError(req.full_url, code, msg, hdrs, fp)
Oct 07 00:05:10 urllib.error.HTTPError: HTTP Error 404: 

cc @pytorch/team-audio-core @mthrok

[svekars]

should we close this issue @malfet ?

[malfet]

I'm not sure if data is available. I.e. there is a workaround, but it points to an older training set

[nateanl]

the newest checkpoint will throw the 404 error, but when I fetch it without .pth extension via wget, it can be downloaded..

wget https://api.ngc.nvidia.com/v2/models/nvidia/waveglow_ckpt_fp32/versions/19.09.0/files/nvidia_waveglowpyt_fp32_20190427

[malfet]

Yeah, it became available again, but wasn't when the issue was filed, adjusting the code accordingly