Commit 55da605a authored by Phil Wang's avatar Phil Wang
Browse files

default back to distributed friendly soundstream, lower default learning rate... 

default back to distributed friendly soundstream, lower default learning rate on soundstream trainer given feedback
parent 47ee1c39

audiolm_pytorch/soundstream.py

+1 −1
Original line number Diff line number Diff line
@@ -462,7 +462,7 @@ class SoundStream(nn.Module): 
        attn_window_size = 128,

        attn_dim_head = 64,

        attn_heads = 8,

        use_complex_stft_discriminator = True

        use_complex_stft_discriminator = False

    ):

        super().__init__()

        self.target_sample_hz = target_sample_hz # for resampling on the fly

audiolm_pytorch/trainer.py

+1 −1
Original line number Diff line number Diff line
@@ -116,7 +116,7 @@ class SoundStreamTrainer(nn.Module): 
        batch_size,

        data_max_length = None,

        folder,

        lr = 3e-4,

        lr = 2e-4,

        grad_accum_every = 4,

        wd = 0.,

        max_grad_norm = 0.5,

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 = '0.7.7',

  version = '0.7.8',

  license='MIT',

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

  author = 'Phil Wang',