search

facebookresearch / omnivore

Omnivore: A Single Model for Many Visual Modalities
Other
559 stars 39 forks source link

Fine-tuning script #15

Closed ZhangYuanhan-AI closed 2 years ago

ZhangYuanhan-AI commented 2 years ago

Hello,

Thanks for this awesome work for multi-modality pre-training!

Do you have a plan to release the fine-tuning script?

Thanks

ZhangYuanhan-AI commented 2 years ago

In addition, what's the batch size for the Fine-tunning of image classification tasks. Thanks!

mannatsingh commented 2 years ago

Hi @Davidzhangyuanhan , glad you find our work useful! Our plan is to release the training code, including the fine tuning script, in June!

The global batch size we used for fine tuning on image classification tasks was 512.

ZhangYuanhan-AI commented 2 years ago

Thanks!