"is_vqa"=True in preprocessor? Is this correct?
#2
by
RainbowMan1
- opened
Deplot is a chart to table/text model. There is no need to add header text during preprocessing. I think "is_vqa" should be False in preprocessor_config.json.
Hey, we added the header "Generate underlying data table of the figure below:" during training. But in practice I've tried with or without this header, it should work similarly.
Could you provide details of lr and hyperparameters used to train this model?