Commit 52cfee78 authored by Phil Wang's avatar Phil Wang
Browse files

soundstream trainer should auto resample the input audio to the correct... 

soundstream trainer should auto resample the input audio to the correct sampling hz set on SoundStream
parent 6289eb0b

README.md

+9 −0
Original line number Diff line number Diff line
@@ -153,6 +153,15 @@ loss = train_wrapper( 
)



loss.backward()



# after a lot of training



mock_coarse_token_ids = torch.randint(0, 1024, (1, 128, 3))



fine_token_ids = train_wrapper.generate(

    coarse_token_ids = mock_coarse_token_ids

) # (1, 128, 5)



```



All together now

audiolm_pytorch/trainer.py

+1 −0
Original line number Diff line number Diff line
@@ -114,6 +114,7 @@ class SoundStreamTrainer(nn.Module): 
        self.ds = SoundDataset(

            folder,

            max_length = data_max_length,

            target_sample_hz = soundstream.target_sample_hz,

            seq_len_multiple_of = soundstream.seq_len_multiple_of

        )

setup.py

+1 −1
Original line number Diff line number Diff line
@@ -3,7 +3,7 @@ from setuptools import setup, find_packages 
setup(

  name = 'audiolm-pytorch',

  packages = find_packages(exclude=[]),

  version = '0.0.46',

  version = '0.0.47',

  license='MIT',

  description = 'AudioLM - Language Modeling Approach to Audio Generation from Google Research - Pytorch',

  author = 'Phil Wang',