File size: 171 Bytes
138108b
 
 
9c415fb
430d7ff
1
2
3
4
5
6
---
license: mit
---
This is the feature alignment pre-training work to train only only the multi-modal projector.
"Predict" paragraph given caption, ocr and image token