zjysteven/VLM-Visualizer

Issues

• Will you do this for siglip the new LLaVA as well?

  #5 opened 25 days ago by trinhvg
  3

• Why are attention from CLIP used for overlayed visualization?

  #3 opened 25 days ago by josejhlee
  3

• Attention for <s> or </s>?

  #7 opened 25 days ago by zifuwan
  1

• Visual Attention Visualization

  #6 opened a month ago by zifuwan
  2

• 'CLIPVisionTower' object has no attribute 'image_attentions'

  #4 opened 2 months ago by zli999
  4

• ValueError: The number of FixedLocator locations (77), usually from a call to set_ticks, does not match the number of labels (78).

  #2 opened 3 months ago by jdg900
  4

• Why is the attention over image transposed? It looks more correct without transposing it.

  #1 opened 3 months ago by ntlm1686
  2