Also known as: GPT
Generative Pretrained Transformer (GPT) — тип большой языковой модели на основе архитектуры Transformer, обученный генерировать текст путем предсказания токенов. Сначала он предварительно обучается на крупных текстовых корпусах, а затем может адаптироваться для диалога, следования инструкциям или специализированных задач.
A large language model based on the transformer architecture that generates text. It is first pretrained to predict the next token in texts (a token is typically a word, subword, or punctuation). After their pretraining, GPT models can generate human-like text by repeatedly predicting the token that they would expect to follow. GPT models are usually also fine-tuned, for example with reinforcement learning from human feedback to reduce hallucination or harmful behaviour, or to format the output in a conversationnal format.
A family of Transformer-based large language models developed by OpenAI. GPT variants can apply to multiple modalities, including: - image generation (for example, ImageGPT) - text-to-image generation (for example, DALL-E).