Downloads · 30 days
0
jimmycarter/imagenet1k-clip-big-g-embeds
imagenet1k-clip-big-g-embeds is a text-to-image model from jimmycarter. Use it when you need an image from a text prompt. The card lists the license as openrail.
This is a tensor with dimensions (1001, 42, 77, 1280). 0 through 999 are imagnet classes, while 1000 is the unconditional class (with the same embedding 42 times). The embeddings are the unsquished, final layernormed…
Downloads · 30 days
0
Access
Public
Updated Apr 9, 2023
Repo size
16.6 GB
Likes
0
Public
Click a slice to open those files.
.pt16.6 GB · 100%
From the Hugging Face model README
This is a tensor with dimensions (1001, 42, 77, 1280). 0 through 999 are imagnet classes, while 1000 is the unconditional class (with the same embedding 42 times). The embeddings are the unsquished, final layernormed output of the CLIP text tower for CLIP big G.
The embeddings were generated from the captions and functions contained in imagenet_zeroshot_data.py.