This project demonstrates the end-to-end workflow of a NLP project with Noronha, using a fit tuned LLM named BERT to make intent recognition.
All steps required for running this example can be found in its script.
The dataset used here is ATIS, a standard benchmark dataset widely used as an intent classification.
Those are the key features demonstrated in this example:
- Structured datasets
- Project building
- In-Notebook shortcuts
- Training and deploying
- Routed inference requests
- Fit training of LLM
This example may be reused as a template in other Machine Learning projects by changing some code snippets in these areas:
- Project setup and CLI actions
- Model files and dataset files definition
- Names, descriptions and metadata
- Training notebook
- Parameter injection cell
- Dataset loading cell
- Training cell
- Inference notebook
- Model loading cell
- Prediction function
-
For further commands and usage options, see the CLI reference.
-
For a better understanding of the relationships between entities (models, versions, trainings, etc...) see the data model guide.
-
For running in a more robust configuration set-up or customizing the framework's behaviour, see the configuration manual.