image tokens as input #91
Description
If the input consists of a class label together with partial image tokens, can the pretrained model directly generate the complete image, or would it require retraining or fine-tuning?
{{ message }}
If the input consists of a class label together with partial image tokens, can the pretrained model directly generate the complete image, or would it require retraining or fine-tuning?