train-CLIP How to use clip on chinese dataset?

How to use clip on chinese dataset? Should I change txt_encoder pretrain model with a chinese version?

Feb 18 '22 02:02 zhouwei5113

Exactly! I think that's the only necessary change. Let me know how it goes :)

Feb 18 '22 08:02 Zasder3

I found a default learning rate 3e-3 when using train_finetune.py, which is a suggested learning rate for both image and text encoder, right? @Zasder3

Feb 18 '22 09:02 zhouwei5113

Exactly! I think that's the only necessary change. Let me know how it goes :)

Training on chinese dataset is very difficult to converge...

Mar 03 '22 02:03 zhouwei5113

Bit late to this! An lr that I use frequently is 1e-4, that or something in that family typically gives good results.

Hopefully future users will be able to benefit from your experiments.

Mar 03 '22 07:03 Zasder3

@zhouwei5113 @Zasder3 Hi, maybe you can refer to this repo! https://github.com/OFA-Sys/Chinese-CLIP

Nov 18 '22 15:11 yangapku