dit-base-224-p16-500k-62d53a.pth, dit-large-224-p16-500k-d7a2fb.pth models

[ol-loginov]

In DiT readme (https://github.com/microsoft/unilm/blob/master/dit/README.md) there are links to checkpoints and fine-tunes:

March 2022: release pre-trained checkpoints and fine-tuning checkpoints & codes (DiT-base and DiT-large)

1. Fine-tunes are ok - I can load it into "object_detection" models.
   But pre-trained checkpoints - what model are they for? Their size is twice larger than any object detection model

2. What checkpoints should I take to fine-tune "object_detection"?