About VQAv2 and COCO2014

[OmegatoAlpha1]

HI, I'm trying to train this model, but I'm not sure which parts of VQAv2 shoud be downloaded to satisfy the 83K data, cound you please show the link to download it? By the way, I found that VQAv2 uses coco2014, should I use coco 2014 for the VQAv2 or just using coco2017?

[SY-Xuan]

You can download our converted VQAv2 format at https://huggingface.co/datasets/SY-Xuan/Pink_sft. The differences between coco2014 and coco2017 are train/val split and filename. You can convert the filename manually.

[OmegatoAlpha1]

Thank you for anwering the question, best wish to your works! Shiyu Xuan ***@***.***> 于 2024年11月12日周二 18:24写道：

…

You can download our converted VQAv2 format at https://huggingface.co/datasets/SY-Xuan/Pink_sft. The differences between coco2014 and coco2017 are train/val split and filename. You can convert the filename manually. — Reply to this email directly, view it on GitHub <#10 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A3KYQQ5U5HXGPWCXBUTRROT2AHJNZAVCNFSM6AAAAABQJKVFUCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINZQGE2DQMBZGA> . You are receiving this because you modified the open/close state.Message ID: ***@***.***>