[Bug]: Slow captioning, 4.5 s/it

[PsilAxis]

What happened?

Why is captioning so slow on OneTrainer? i am using WD14 VIT V2, and its captioning speed is 4.5 s/it. captioning 1.7k images is expected to take over 2 hours.

when using other programs like kohya for autocaptioning, with WD14 VIT V2, captioning is done almost instantly. 1.7k is done in a few seconds, at batch size 1.

What did you expect would happen?

Faster captioning like with other programs.

Relevant log output

7%|▋         | 129/1726 [09:36<1:57:21,  4.41s/it]
  8%|▊         | 130/1726 [09:40<1:56:25,  4.38s/it]

Output of pip freeze

No response