About evaluation data on chart-to-text and chart-to-table

[zhangliang-04]

Hi,
Thanks for the open sourcing! I want to reproduce the performance of chart-to-text and chart-to-table, and find that the evaluation scripts are released here and here in this repo but the corresponding JSON files that contain the data and prompts are not there. Could you release them?
By the way, I noticed that Chartllama may be further separately trained with Pew and Statista for chart-to-text evaluation [1]. Did you do similar things for chart-to-table? Is the current released model able to reproduce the performance, or does it have to be further trained on the downstream datasets?
Thanks again for your great work. I'd appreciate it if you could help me!

[anshumankmr]

Hey did you have any luck running this yourself ? I am also hoping to use this for making graphs from structured datasets.