Building and Deploying Conversational AI Models Using the NVIDIA Transfer Learning Toolkit

Originally published at:

Conversational AI is a set of technologies enabling human-like interactions between humans and devices based on the most natural interfaces for us: speech and natural language. Systems based on conversational AI can understand commands by recognizing speech and text, translating on-the-fly between different languages, understanding our intents, and responding in a way that mimics human…

Hello, for the fine tuning part of this exercise, is the data available? When I stepped through the example I hit an error there. Also, for the KEY bash variable, i set this to my own user specified value, but I kept getting errors telling me I had an incorrect format. In a different TLT example, i found the authors set key to “tlt_encode” and then I was able to make progress.

Hi @dvanstee ,

Generally we/NVIDIA don’t distribute data(sets) that we don’t own. Still, for the Text Classification task, TLT supports two public datasets out-of-the-box (SST-2 and IMBD) that you can download on your own. Then simply run dataset_convert script with a proper dataset name (please refer to TLT Text Classification user guide). Those datasets can be used for both training (from scratch) and/or fine-tuning.

When it comes to models used as a starting point for fine-tuning, I guess you downloaded them from NGC. If so, then please follow the instructions regarding usage of a given model (and key in particular) provided in the associated NGC Model Card.

Hope it helps,