At ICML 2019, PyTorch released PyTorch Hub, a repository of pre-trained models designed specifically for reproducible research. PyTorch is one of the most widely used deep learning frameworks by researchers and developers. PyTorch 1.0, announced by Facebook in 2018, is a deep learning framework that powers numerous products and services at scale by merging the best of both worlds – the distributed and native performance found in Caffe2 and the flexibility for rapid development found in the existing PyTorch framework. At a high level, PyTorch is a Python package that provides high level features such as tensor computation with strong GPU acceleration.
PyTorch Hub is a simple API and workflow the provides the basic building blocks for improving machine learning research reproducibility. PyTorch Hub consists of a pre-trained model repository designed specifically to facilitate research reproducibility. What this means is that researchers in various fields can easily discover each other’s research, leverage it as a baseline and build new cutting edge research from there. Users can discover available models on pytorch.org/hub along with a rich set of metadata, preprocessing steps and the background research. Current Hub contributions from Nvidia include Text-to-Speech models based on Tacotron2 and WaveGlow with more to come.
Current models contributed by Nvidia on PyTorch Hub today:
- Tacotron2 and WaveGlow: The Tacotron2 and WaveGlow models produced by NVIDIA research team provide state of the art techniques that offer a text-to-speech system which enables user to synthesise a natural sounding speech from raw transcripts without any additional prosody information. This text-to-speech (TTS) system is a combination of two neural network models: a modified Tacotron2 model from the Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions paper and a flow-based neural network model from the WaveGlow: A Flow-based Generative Network for Speech Synthesis paper.
To access this on Torch Hub, use the command
torch.hub.load(‘nvidia/DeepLearningExamples’, ‘nvidia_tacotron2’, pretrained=True)
torch.hub.load(‘nvidia/DeepLearningExamples’, ‘nvidia_waveglow’, pretrained=True)
NVIDIA strives to bring innovation and flexibility to the deep learning developer community. We are further investing in our engineering efforts and working together to empower and engage with the PyTorch developer community. The full article is available on pytorch.org/blog and you can learn more about all of the pre-built examples built by NVIDIA with state of the art accuracy for various frameworks here.
We are enabling more experiences that matter the most to our developer community. Get notified on NVIDIA and the larger PyTorch ecosystem enabling the next generation of powerful AI use-cases, and more news to your inbox. Subscribe to developer newsletter to be notified on the latest product updates. If you are a developer or researcher looking for technical resources on deep learning, check out NVIDIA Developer Blogs to get started today!