Use GradCam in ViT with visualization problem, the attention map not show

[Like-wind-fly]

python tools/vis_cam.py 1.jpg D:\code\lesion-classification\Awesome-Backbones\logs\VisionTransformer\2024-12-31-15-32-22\vit_base_p16_224.py --save-path logs\VisionDTransformer\2024-12-31-15-32-22\VisionTransformer_gradcam_wild.jpg --vit-like --target-layers backbone.layers[-1].ln1

Could you give me a solution or explanation, thank you so much.

[Shirley-xyy]

I encountered this issue with ConvNeXt as well, and when I switched to other layers, the result showed vertical lines.