mingiglai issues

Repositories
Issues
Comments

Results 2 issues of


                                            mingiglai

Issue with duplicated data in DocDownStream1.0

Hi, I notice that the train.jsonl file downloaded from huggingface under DocDownStream1.0 contains a lot of duplicate data depending on the data source. For example, each data in DocVQA is...

For generated solutions with non-python code, how are the solution COT generated by the model cleaned in the evaluation process?

I can't locate the core logic for cleaning the model's generated solution from non-python code in the codes within evaluation directory. Currently, i can only see how dataset ground truth...