Prompt Study: How Transformer Works: Past Picture Warehouse Series
2023-02-01 22:56:07
対象年齢：全年齢
Source URL:
https://www.slideshare.net/ShotaImai3/chatgpt-254863623/ShotaImai3/chatgpt-254863623
Diffusion model seems to use an open source dataset called CLIP.
CLIP also uses the Transformer system, so I think the mechanism is the same.
However, it seems that there are hundreds of billions of parameters in the dataset, and hundreds of thousands of dimensions, so it is impossible to search the full text.
Therefore, it seems that they are doing a detailed search after grouping images with similar properties together as a category, narrowing down the category roughly.
https://rom1504.github.io/clip-retrieval/?back=https%3A%2F%2Fknn.laion.ai
ログインするとプロンプトなどがチェックできます
※ 作品によっては掲載されていないことがあります
コメント