AttributeError: module 'torch' has no attribute 'bucketize'

[loretoparisi]

I get the following error:

root@75adae8f35d1:/app# python3 synthesize.py --step 300000
|{DH AH0 N EY1 SH AH0 N Z T UH1 R IH2 Z AH0 M M IH1 N AH0 S T ER0 HH AE1 Z AO1 L S OW0 EH0 N K ER1 IH0 JH D AO2 S T R EY1 L Y AH0 N Z T UW1 T EY1 K DH EH1 R HH AA1 L AH0 D EY2 Z W IH0 DH IH1 N DH AH0 K AH1 N T R IY0 DH IH1 S Y IH1 R} |
Traceback (most recent call last):
  File "synthesize.py", line 94, in <module>
    synthesize(model, text, sentence, prefix='step_{}'.format(args.step))
  File "synthesize.py", line 48, in synthesize
    mel, mel_postnet, duration_output, f0_output, energy_output = model(text, src_pos)
  File "/usr/local/lib/python3.7/site-packages/torch/nn/modules/module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "/usr/local/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 143, in forward
    return self.module(*inputs, **kwargs)
  File "/usr/local/lib/python3.7/site-packages/torch/nn/modules/module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "/app/fastspeech2.py", line 33, in forward
    encoder_output, d_target, p_target, e_target, max_length)
  File "/usr/local/lib/python3.7/site-packages/torch/nn/modules/module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "/app/modules.py", line 47, in forward
    pitch_embedding = self.pitch_embedding(torch.bucketize(pitch_prediction, self.pitch_bins))
AttributeError: module 'torch' has no attribute 'bucketize'

I'm running in CPU, and I had to modify get_FastSpeech2 like this:

def get_FastSpeech2(num):
    checkpoint_path = os.path.join(hp.checkpoint_path, "checkpoint_{}.pth.tar".format(num))
    model = nn.DataParallel(FastSpeech2())
    if torch.cuda.is_available():
        model.load_state_dict(torch.load(checkpoint_path)['model'])
    else:
        model.load_state_dict(torch.load(checkpoint_path, map_location=torch.device('cpu'))['model'])
    model.requires_grad = False
    model.eval()
    return model

to set map_location to cpu device.

[ming024]

@loretoparisi What is your PyTorch version? This functionality is newly introduced in PyTorch 1.6, which is still not stable yet.

[loretoparisi]

@ming024 got it, I was using 1.5.0 but someone told me I have to install

RUN pip install --pre torch==1.6.0.dev20200428 torchvision -f https://download.pytorch.org/whl/nightly/cu101/torch_nightly.html

Thank you, I will try this way.

[loretoparisi]

@ming024 I have fixed the error using 1.6.0dev, but now I'm getting

Traceback (most recent call last):
  File "synthesize.py", line 94, in <module>
    synthesize(model, text, sentence, prefix='step_{}'.format(args.step))
  File "synthesize.py", line 61, in synthesize
    Audio.tools.inv_mel_spec(mel_postnet, os.path.join(hp.test_path, '{}_griffin_lim_{}.wav'.format(prefix, sentence)))
  File "/app/audio/tools.py", line 63, in inv_mel_spec
    spec_from_mel[:, :, :-1]), _stft.stft_fn, griffin_iters)
  File "/app/audio/audio_processing.py", line 74, in griffin_lim
    _, angles = stft_fn.transform(signal)
  File "/app/audio/stft.py", line 66, in transform
    input_data.cuda(),
  File "/usr/local/lib/python3.7/site-packages/torch/cuda/__init__.py", line 150, in _lazy_init
    _check_driver()
  File "/usr/local/lib/python3.7/site-packages/torch/cuda/__init__.py", line 54, in _check_driver
    http://www.nvidia.com/Download/index.aspx""")
AssertionError: 
Found no NVIDIA driver on your system. Please check that you
have an NVIDIA GPU and installed a driver from
http://www.nvidia.com/Download/index.aspx

I'm using CPU.

[loretoparisi]

[UPDATE]
Ok I have fixed the previous one in sftp.py doing in def transform(self, input_data): like:

if torch.cuda.is_available():
            forward_transform = F.conv1d(
                input_data.cuda(),
                Variable(self.forward_basis, requires_grad=False).cuda(),
                stride=self.hop_length,
                padding=0).cpu()
        else:
            forward_transform = F.conv1d(
                input_data,
                Variable(self.forward_basis, requires_grad=False),
                stride=self.hop_length,
                padding=0).cpu()

the problem is that it seems that the waveglov

root@b832c651a4bc:/app# python synthesize.py --step 300000
|{DH AH0 N EY1 SH AH0 N Z T UH1 R IH2 Z AH0 M M IH1 N AH0 S T ER0 HH AE1 Z AO1 L S OW0 EH0 N K ER1 IH0 JH D AO2 S T R EY1 L Y AH0 N Z T UW1 T EY1 K DH EH1 R HH AA1 L AH0 D EY2 Z W IH0 DH IH1 N DH AH0 K AH1 N T R IY0 DH IH1 S Y IH1 R} |
Using cache found in /root/.cache/torch/hub/seungwonpark_melgan_master
Traceback (most recent call last):
  File "synthesize.py", line 94, in <module>
    synthesize(model, text, sentence, prefix='step_{}'.format(args.step))
  File "synthesize.py", line 64, in synthesize
    melgan = utils.get_melgan()
  File "/app/utils.py", line 132, in get_melgan
    melgan = torch.hub.load('seungwonpark/melgan', 'melgan')
  File "/usr/local/lib/python3.7/site-packages/torch/hub.py", line 362, in load
    model = entry(*args, **kwargs)
  File "/root/.cache/torch/hub/seungwonpark_melgan_master/hubconf.py", line 19, in melgan
    progress=progress)
  File "/usr/local/lib/python3.7/site-packages/torch/hub.py", line 502, in load_state_dict_from_url
    return torch.load(cached_file, map_location=map_location)
  File "/usr/local/lib/python3.7/site-packages/torch/serialization.py", line 580, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/usr/local/lib/python3.7/site-packages/torch/serialization.py", line 760, in _legacy_load
    result = unpickler.load()
  File "/usr/local/lib/python3.7/site-packages/torch/serialization.py", line 716, in persistent_load
    deserialized_objects[root_key] = restore_location(obj, location)
  File "/usr/local/lib/python3.7/site-packages/torch/serialization.py", line 174, in default_restore_location
    result = fn(storage, location)
  File "/usr/local/lib/python3.7/site-packages/torch/serialization.py", line 150, in _cuda_deserialize
    device = validate_cuda_device(location)
  File "/usr/local/lib/python3.7/site-packages/torch/serialization.py", line 134, in validate_cuda_device
    raise RuntimeError('Attempting to deserialize object on a CUDA '
RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

and If I do in utils.py:

def get_melgan():
    if torch.cuda.is_available():
        melgan = torch.hub.load('seungwonpark/melgan', 'melgan')
    else:
        # nvidia's waveglow models explicitly don't work on cpu.
        melgan = torch.hub.load('seungwonpark/melgan', 'melgan', map_location=torch.device('cpu'))
    melgan.eval()
    return melgan

I now get

root@b832c651a4bc:/app# python synthesize.py --step 300000
|{DH AH0 N EY1 SH AH0 N Z T UH1 R IH2 Z AH0 M M IH1 N AH0 S T ER0 HH AE1 Z AO1 L S OW0 EH0 N K ER1 IH0 JH D AO2 S T R EY1 L Y AH0 N Z T UW1 T EY1 K DH EH1 R HH AA1 L AH0 D EY2 Z W IH0 DH IH1 N DH AH0 K AH1 N T R IY0 DH IH1 S Y IH1 R} |
Using cache found in /root/.cache/torch/hub/seungwonpark_melgan_master
Traceback (most recent call last):
  File "synthesize.py", line 94, in <module>
    synthesize(model, text, sentence, prefix='step_{}'.format(args.step))
  File "synthesize.py", line 64, in synthesize
    melgan = utils.get_melgan()
  File "/app/utils.py", line 133, in get_melgan
    melgan = torch.hub.load('seungwonpark/melgan', 'melgan', map_location='cpu')
  File "/usr/local/lib/python3.7/site-packages/torch/hub.py", line 362, in load
    model = entry(*args, **kwargs)
TypeError: melgan() got an unexpected keyword argument 'map_location'

So an alternative approach to pass map_location was this one

def get_melgan():
    if torch.cuda.is_available():
        melgan = torch.hub.load('seungwonpark/melgan', 'melgan')
        melgan.eval()
    else:
        # nvidia's waveglow models explicitly don't work on cpu.
        melgan = torch.hub.load('seungwonpark/melgan', 'melgan', pretrained=False)
        checkpoint = torch.hub.load_state_dict_from_url('file:///root/.cache/torch/hub/checkpoints/nvidia_tacotron2_LJ11_epoch6400.pt ', map_location="cpu")
        #state_dict = {key.replace("module.", ""): value for key, value in checkpoint["state_dict"].items()}
        melgan.load_state_dict(checkpoint)
        melgan.eval()
    return melgan

but it does not unwrap correctly the checkpoint's state dict.

[ming024]

@loretoparisi it is because in the MelGAN implementation the weight normalization parameters are removed when the model is set to inference mode.
https://github.com/seungwonpark/melgan/blob/aca59909f6dd028ec808f987b154535a7ca3400c/hubconf.py#L22

You can see this colab for a simple solution.

[ming024]

closed #7