
Amazon SageMaker JumpStart now enables zero-shot text classification, allowing pre-trained models to classify text into unseen categories without fine-tuning through UI deployment or Python SDK integration.


<div><p>This article explains how to perform zero-shot text classification using Amazon SageMaker JumpStart, enabling text classification without explicit training on target classes.</p><ul><li>Zero-shot classification classifies unseen text categories without model fine-tuning</li><li>SageMaker JumpStart provides pre-trained models like facebook-bart-large-mnli for deployment</li><li>Deploy models via SageMaker Studio UI or programmatically using Python SDK</li><li>Supports real-time inference through SageMaker endpoints for single predictions</li><li>Supports batch inference using SageMaker batch transform for large-scale processing</li><li>Uses natural language inference framework to convert classification to entailment tasks</li><li>Example uses New Year's resolutions dataset with nine classification categories</li></ul><p>SageMaker JumpStart simplifies zero-shot text classification deployment, offering both UI and SDK approaches for real-time and batch inference without requiring model retraining.</p></div>


Related articles