bytedance / 1d-tokenizer

This repo contains the code for our paper An Image is Worth 32 Tokens for Reconstruction and Generation
Apache License 2.0
404 stars 16 forks source link

Request for training code #4

Closed ChuxiJ closed 1 month ago

ChuxiJ commented 3 months ago

image

Great work!

Want to reproduce the results of your paper and do some experiments, can we get the training code?

Maybe we can replace the vq with FSQ and increase the codebook size from 4096 into bigger number so we can find a token level with optimal compact size and reconstruction quality.

Thanks!

cornettoyu commented 3 months ago

Hi,

Improving TiTok with FSQ or larger codebook sounds promising and interesting. The training code is still waiting for internal review and approval. Feel free to let me know if you have any questions regarding technical details before we release the full codebase.

numb3r3 commented 3 months ago

Congrats on your remarkable work. Looking forward to the training codes.

zimenglan-sysu-512 commented 3 months ago

will release the training code?

yyyouy commented 3 months ago

Thank you for your excellent work. I would like to inquire about the expected time for internal review and approval.

cornettoyu commented 1 month ago

Hi,

Thank you all for your interests in our work and patience. We have released the two-stage training code and feel free to check it.