Open lassloss opened 7 months ago
Sorry to bother you, are the weights provided in vit_model.py about vit_base_patch16_224 ("https://pan.baidu.com/s/1zqb08naP0RPqqfSXfkB2EA") pre-trained on ImageNet-1k only ? Or does it also use ImageNet-21k ?
Sorry to bother you, are the weights provided in vit_model.py about vit_base_patch16_224 ("https://pan.baidu.com/s/1zqb08naP0RPqqfSXfkB2EA") pre-trained on ImageNet-1k only ? Or does it also use ImageNet-21k ?
I'm not sure whether the models in Baidu Cloud Disk are pre-trained on imagenet21k. But I didn't find the corresponding model in https://github.com/google-research/vision_transformer.
Thank you for your reply.
Could you provide a pretrained vit on imagenet 1k? I can only find a model that has been fine-tuned on imagenet 1k after training on imagenet 21k.