search

lucidrains / CoCa-pytorch

Implementation of CoCa, Contrastive Captioners are Image-Text Foundation Models, in Pytorch
MIT License
1.04k stars 88 forks source link

Reproduce Problem #19

Closed Camellia-tx closed 10 months ago

Camellia-tx commented 10 months ago

Thanks for this repo. I'm using my own dataset for pre-training via CoCa, and I found that the contrastive loss output from each batch is basically unchanged, is it the contrastive_label that needs to be changed or is there some other place where I need to make corresponding changes? Thank you!!!!

lucidrains commented 10 months ago

the code in this repo has been used to reproduce an open sourced model already.

Camellia-tx commented 10 months ago

I apologize for only replying now, thank you so much for your contribution!

在 2023-10-27 23:30:19，"Phil Wang" @.***> 写道：

Closed #19 as completed.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>