Code for reproducing evaluation results on ScienceQA

[TJKlein]

Hi, For reasons of reproducibility, it would be great if you provided source code to reproduce the results on ScienceQA. Thanks.

[gaopengpjlab]

Sure, we are going to release all training/inference codes for all results reported in paper.

[gaopengpjlab]

@TJKlein Please the following multimodal inference code for image captioning. The code structure is very similar as ScienceQA.

https://huggingface.co/spaces/csuhan/LLaMA-Adapter/tree/main .

[TJKlein]

@TJKlein Please the following multimodal inference code for image captioning. The code structure is very similar as ScienceQA.

https://huggingface.co/spaces/csuhan/LLaMA-Adapter/tree/main .

Thanks for pointing at it. But I guess I would wait for the release of your scripts.

[yushuinanrong]

+1

[basteran]

Hi I am interested too in the integration of the visual adapter. When do you think the code will be released?

[ZrrSkywalker]

Thanks for your interest and waiting :). We are organizing the multi-modal code and will release it in one or two weeks.

[basteran]

Thank you for your quick response!

[metemadi]

hi there! any updates on this? would love to take a crack at fine-tuning on a visual instruction dataset (not just image-caption as in your amazing work in v2!).

[dhyani15]

Hi, just wanted to check for any updates on this?