CLIP

CLIP (Contrastive Language-Image Pre-Training) is a neural network trained on a variety of (image, text) pairs. It can be instructed in natural language to predict the most relevant text snippet, given an image, without directly optimizing for the task, similarly to the zero-shot capabilities of GPT-2 and 3.

homepage: https://github.com/openai/CLIP

version	versionsuffix	toolchain
20230220	-CUDA-11.7.0	foss/2022a

(quick links: (all) - 0 - a - b - c - d - e - f - g - h - i - j - k - l - m - n - o - p - q - r - s - t - u - v - w - x - y - z)