This past year, NVIDIA announced several major breakthroughs in conversational AI for building and deploying automatic speech recognition (ASR), natural language processing (NLP), and text-to-speech (TTS) applications.
To get developers started with some quick examples in a cloud GPU-accelerated environment, NVIDIA Deep Learning Institute (DLI) is offering three fast, free, self-paced courses.
What will you learn?
These instructional DLI courses give developers a taste of how to use modern tools to quickly create conversational AI and NLP GPU-accelerated applications. Learning objectives include:
- Train a Text Classification Model Using TAO Toolkit
- Train and fine-tune a BERT text classification model on the SST-2 dataset.
- Run evaluation and inference on the model.
- Export the model to ONNX format or Riva format for deployment.
- Deploy a Text Classification Model Using Riva
- Use Riva ServiceMaker to take a TAO-exported Riva model and convert it for final deployment.
- Deploy the model(s) locally on the Riva Server.
- Send inference requests from a demo client using Riva API bindings.
- Riva Speech API Demo
- Send audio to an ASR model and receive back text.
- Use NLP models to transform text, classify text, and classify tokens.
- Send text to a TTS model and receive back audio.
Upon course completion, developers will be familiar with:
- How to train, infer, and export a text classification model using NVIDIA TAO Toolkit on NVIDIA GPUs.
- How to deploy a text classification model using NVIDIA Riva on NVIDIA GPUs.
- How to construct requests to an NVIDIA Riva Speech server from a sample client.
Why is text classification useful?
Text classification answers the question: Which category does this bit of text belong in? For example, if you want to know whether a movie review is positive or negative, you can use two categories to build a sentiment analysis project.
Take this one step further, and classify sentences or documents by topic using several categories. In both use cases, you start with a pre-trained language model and then “train” a classifier using example classified text to create our text classification project.
Granted, text classification is just one of many NLP tasks that uses a pre-trained language model to understand written language. Once developers try NVIDIA TAO Toolkit and NVIDIA Riva to train and deploy text classification projects, they will be in a position to extend that experience to additional NLP tasks, such as named entity recognition (NER) and question answering.
How does the NVIDIA Riva Speech API work?
The Riva Speech API server exposes a simple API for performing speech recognition, speech synthesis, and a variety of NLP inferences. In this course, developers use Python examples to run several of these API calls from within a Riva sample client. The server is prepopulated with ASR, NLP, and TTS models. These built-in models allow developers to test several conversational AI components quickly with ease.
Start learning about NLP and conversational AI
- Train a Text Classification Model Using TAO Toolkit (60 minutes)
- Deploy a Text Classification Model Using Riva (30 minutes)
- Riva Speech API Demo (30 minutes)