Where is the "EVA02_CLIP_L_336_psz14to16_s6B.pt"?

[becauseofAI]

Is this EVA02_CLIP_L_336_psz14to16_s6B.pt necessary for training, validation, or inference? If so, where is the download link? thx @shenyunhang

[lxw919]

You can run git lfs install git clone https://huggingface.co/QuanSun/EVA-CLIP models/QuanSun/EVA-CLIP/

and lfs need be installed before,the process is a little slow.

[shenyunhang]

EVA02_CLIP_L_336_psz14to16_s6B.pt is only used in training.

For validation and inference, the trained-model path should overwrite train.init_checkpoint to initialize the model.

[becauseofAI]

EVA02_CLIP_L_336_psz14to16_s6B.pt is only used in training.

For validation and inference, the trained-model path should overwrite train.init_checkpoint to initialize the model.

So can you provide its download link? @shenyunhang

[shenyunhang]

You can download the original model from here Then covert it by running:

python3 tools/eva_interpolate_patch_14to16.py --input models/QuanSun/EVA-CLIP/EVA02_CLIP_L_336_psz14_s6B.pt --output models/QuanSun/EVA-CLIP/EVA02_CLIP_L_336_psz14to16_s6B.pt --image_size 336

To prepare other backbones, please follow those steps.