CLIP-self-attention-visualization
CLIP-self-attention-visualization copied to clipboard
Plotting heatmaps with the self-attention of the [CLS] tokens in the last layer.
Results
1
CLIP-self-attention-visualization issues
Sort by
recently updated
recently updated
newest added
When I changed the pretrained model fro ViT-B/32 to ViT-B/16 and ViT-L/14, I could not get such an attention map. Do you have the same results?