T5 (Text-to-Text Transfer Transformer)

par Google

AI Model

Faible confiance

T5 is a transformer-based language model developed by Google Research that reframes all natural language processing (NLP) tasks into a unified text-to-text format. This means both input and output are always text strings, allowing the model to handle tasks like translation, summarization, question answering, and classification using the same architecture. T5 is pre-trained on a large corpus of web-extracted text (C4) and can be fine-tuned for specific downstream tasks. It is designed to be highly scalable and adaptable, with variants ranging from small (60 million parameters) to XXL (11 billion parameters).