Zeju1997 / oft

Official implementation of "Controlling Text-to-Image Diffusion by Orthogonal Finetuning".
https://oft.wyliu.com/
MIT License
280 stars 14 forks source link

what is the use of prompt_token in eval_ablation.py? #17

Closed DaShenZi721 closed 7 months ago

DaShenZi721 commented 7 months ago

Hello, thanks for releasing this code.

I'm running the Subject-driven Generation task. After setting i=0 and running ./train_dreambooth_oft.sh $i, I got such directory structure

log_oft
├── backpack-0
│ ├── 0
│ │ ├── image_0.png
│ │ ├── image_1.png
│ │ ├── ...
│ │ └── image_7.png
│ ├── 1
│ │ ├── image_0.png
│ │ ├── image_1.png
│ │ ├── ...
│ │ └── image_7.png
│ └── ...
...

Then, I run the evaluation script python eval_ablation.py, but the result is dino mean_similarity nan.

I found line 159 in oft-db/eval_ablation.py if re.search(prompt_token, subfolder): didn't match any of the subfolders.

What is the use of prompt_token? And Is there a problem with my training process? Looking forward to your reply.

DaShenZi721 commented 7 months ago

I see.