yaml configuration of validation and test splits for downstream datasets

[franciszchen]

Hi, thanks for sharing this great project! I am transferring the minigpt-4 to a downstream dataset. The training of downstream dataset is fine, but for the validation/test splits I encountered some problems with the dataloaders, where the vis_processor of the validation/test splits is None. In the dataloader of validation/test splits, my script can access the samples. I think the yaml configuration may be some problem. Here is my configuration for the downstream dataset.

datasets:
  downstream_dataset:
    vis_processor:
      train:
        name: "blip2_image_train"
        image_size: 224
    text_processor:
      train:
        name: "blip_caption"

I am wondering how to modify it for the validation and test splits, as well as the keywords for them. Thx~

[hao1635]

Have you completed the test code？ How to set the yaml configuration and dataset? @franciszchen

[Abhiram4572]

Hello! Did you find out the way for the same? If so, can you please guide by sharing the yaml configuration?