Closed qiuhuiGithub closed 1 year ago
For pretraining dataset, you can refer to Laion400M and COYO-700M which is publicly available.
For pre-training, we have no time to process it, you may can refer to the finetuning scripts which is similar except for the prompt. Or you can directly using the checkpoint we provided which we recommend.
hi, will you released code and dataset for pretrained?