hamidkazemi22/vit-visualization

This is stars and forks stats for /hamidkazemi22/vit-visualization repository. As of 10 May, 2024 this repository has 153 stars and 10 forks.

What do Vision Transformers Learn? A Visual Exploration ViT Models: To visualize the features of the ViT models: PYTHONPATH=. python experiments/it15/vis35.py -l <layer_number> -f <feature_number> -n <network_number> -v <tv_coefficient> For example: PYTHONPATH=. python experiments/it15/vis35.py -l 4 -f 20 -n 35 -v 0.1 Clip Models: To visualize the features of the CLIP models: PYTHONPATH=. python experiments/it15/vis98.py -l <layer_number> -f <feature_number>...