Team Ai
Apppublic

SasikaA073/transformer-explainability

sourceHugging Faceupdated 10mo agoView on Hugging Face
0likes
App README

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

How to run backend

backend

python flask_backend.py --port 5001

frontend

go to frontenf folder npm install npm run dev

Model Information

model_idavailable?backbone only?has registers?input_image_resolutionpatch_sizefinetuned_head_path
vitbasepatch8_224☐☐☐2248-
vitbasepatch16224.augreg2in21kftin1k☑☐☐22416-
vitbasepatch16_224☐☑☐22416-
vitlargepatch16rope224☐☐☐22416-
vithugepatch16_224☐☐☐22416-
vitlargepatch14_224☐☑☐22414-
vithugepatch14_224☐☑☐22414-
deit3smallpatch16224.fbin22kftin1k☑☐☐22416-
deit3basepatch16224.fbin22kftin1k☑☑☐22416-
deit3mediumpatch16224.fbin22kftin1k☑☐☐22416-
deit3largepatch16224.fbin22kftin1k☐☐☐22416-
deit3smallpatch16384.fbin22kftin1k☐☐☐38416-
deit3basepatch16384.fbin22kftin1k☐☐☐38416-
deit3largepatch16384.fbin22kftin1k☐☐☐38416-
vitsmallpatch14_dinov2☐☐☐51814-
vitbasepatch14_dinov2☑☑☐51814vitbasepatch14dinov2in1k_best.pth
vitsmallpatch14reg4dinov2☐☐☑51814-
vitbasepatch14reg4dinov2☑☑☑51814vitbasepatch14reg4dinov2headfinetuned_best.pth
vitbasepatch16dinov3in1k☐☐☐51816-
vitbasepatch16_dinov3☑☑☐51816vitbasepatch16dinov3in1kbest20_epochs.pth
vitbasepatch16dinov3qkvb☐☐☐51816-

TODO:

  • —add multiple pytorch models support
  • —use timm library
  • —choose model from the dropdown list
  • —View clicked image full screen to view finer details

UI Fixes

  • —add inverse mask
  • —change the brush size
  • —add reset mask
  • —need to pasrse the pertubed image and get the explainability map
  • —host on hugging face hub
>>>>>> ced0342 (Prepare for Hugging Face Deployment)