Gretel Navigator - Getting Started
Video description
A walkthrough of a common use case for Gretel Navigator
More Videos
Read the blog post
Transcription
Speaker 1 (00:00):
Hey, this is Alex from Gretel. Today we're going to walk through a really basic example of getting started with tabular lm. Inside the Gretel console, you go ahead and sign in from gretel.ai. Once you're in, go ahead and click on the model playground. This is the place that you interactively prompt the tabular LM model, or even our natural language DPT models, both available. We'll start with a super basic prompt here. So we'll drag in one of the examples. This is creating a mock dataset for users in France. And the workflow that we see is really the ability for you to go through the prompt here, look at the results, tune the results. For example, if you wanted to add a new column, you could go ahead and add a new column. So it's ripping through here, generating a bunch of data. And once you're happy with this, then you have the ability to generate a much larger data set in the hundreds or thousands of columns or even more.
(00:46)
So at that point, we move to a batch mode. So I'm going to go ahead and click Generate more data. This allows me to select the Gretel project for this to exist in. I can choose how many records and what the LM to generate of tabular data, and I bring the prompt in that I just tuned and I'm really happy the results with from the previous step click generate. So what's going to happen here? It will fire up a container in our backend. The container is going to follow a plan and execute architecture based approach. So it's going to take this prompt and divide it up into a set of small problems. The LLM is going to execute on those problems and then return the full size data set. We're looking for whether it's a thousand or a hundred thousand columns for us. Very quickly here we can see it's already done. So happened pretty fast here. The model got fired up. We can see the records. Downloads are available now and the job is completed. So we can go ahead and access our artifact right here and go ahead and download the records, open them, and we are good to go.