From a1c77fb5279c2abdbac3a01bb61b423036df85c5 Mon Sep 17 00:00:00 2001 From: Brandon Thomas Date: Wed, 2 Jun 2021 22:26:55 -0400 Subject: [PATCH] Copying waveglow files @ 2fd4e63 (per collab synthesis.ipynb) --- models/tacotron2/waveglow/README.md | 4 ++-- models/tacotron2/waveglow/convert_model.py | 5 +---- 2 files changed, 3 insertions(+), 6 deletions(-) diff --git a/models/tacotron2/waveglow/README.md b/models/tacotron2/waveglow/README.md index b94648cf8e..89886f00a6 100755 --- a/models/tacotron2/waveglow/README.md +++ b/models/tacotron2/waveglow/README.md @@ -12,7 +12,7 @@ using only a single network, trained using only a single cost function: maximizing the likelihood of the training data, which makes the training procedure simple and stable. -Our [PyTorch] implementation produces audio samples at a rate of 1200 +Our [PyTorch] implementation produces audio samples at a rate of 4850 kHz on an NVIDIA V100 GPU. Mean Opinion Scores show that it delivers audio quality as good as the best publicly available WaveNet implementation. @@ -85,7 +85,7 @@ with fused residual and skip connections. [Glow]: https://blog.openai.com/glow/ [WaveNet]: https://deepmind.com/blog/wavenet-generative-model-raw-audio/ [PyTorch]: http://pytorch.org -[published model]: https://drive.google.com/open?id=1rpK8CzAAirq9sWZhe9nlfvxMF1dRgFbF +[published model]: https://ngc.nvidia.com/catalog/models/nvidia:waveglow_ljs_256channels [mel-spectrograms]: https://drive.google.com/file/d/1g_VXK2lpP9J25dQFhQwx7doWl_p20fXA/view?usp=sharing [LJ Speech Data]: https://keithito.com/LJ-Speech-Dataset [Apex]: https://github.com/nvidia/apex diff --git a/models/tacotron2/waveglow/convert_model.py b/models/tacotron2/waveglow/convert_model.py index 6fdc41e871..32e77e2b20 100644 --- a/models/tacotron2/waveglow/convert_model.py +++ b/models/tacotron2/waveglow/convert_model.py @@ -61,15 +61,12 @@ def update_model(old_model): _update_model_res_skip(old_model, new_model) if hasattr(old_model.WN[0], 'cond_layers'): _update_model_cond(old_model, new_model) - for m in new_model.modules(): - if 'Conv' in str(type(m)) and not hasattr(m, 'padding_mode'): - setattr(m, 'padding_mode', 'zeros') return new_model if __name__ == '__main__': old_model_path = sys.argv[1] new_model_path = sys.argv[2] - model = torch.load(old_model_path, map_location='cpu') + model = torch.load(old_model_path) model['model'] = update_model(model['model']) torch.save(model, new_model_path)