New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

training setdata #15

Open

zi1zi opened this issue Apr 24, 2023 · 1 comment

zi1zi commented Apr 24, 2023

I use LF-Font for training and have the following problem:

Are all the unicode encodings in the given example data/chn/train_chars.json used for training? I think the number of characters is very large. Don’t we have fewer samples?
fewshot-font-generation/LF/phase1_trainer.py

Line 61 in c445f66

ref_imgs = batch["ref_imgs"].cuda()

in this link
ref_imgs = batch["ref_imgs"].cuda(), why there is a ref_imgs field in the batch

Collaborator

8uos commented Jun 16, 2023

Hi,

Yes, we use all the characters in data/chn/train_chars.json for training. As mentioned in our paper, our method uses large number of characters using training phase.
We use a custom collate_fn for dataloader to have a dictionary-formatted batch. See here.

Sorry for the late reply.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment