I notice that there is a validate() function in MultimodalTrainer and you validate the trained model after training. Since the validate() function edits the model with the validation set and returns metrics, what's the difference between it and the MultimodalEditor.edit_dataset function, which also edits the model with the validation set and returns metrics. It seems like that thay play the same role. But I get different values of these metrics when I conduct the two functions.
I notice that there is a validate() function in
MultimodalTrainerand you validate the trained model after training. Since the validate() function edits the model with the validation set and returns metrics, what's the difference between it and theMultimodalEditor.edit_datasetfunction, which also edits the model with the validation set and returns metrics. It seems like that thay play the same role. But I get different values of these metrics when I conduct the two functions.