Datasets:

ArXiv:
License:

Can't find images for `osatlas_ui_tars_cleaned.jsonl`

#2
by rronan-h - opened

In datasets.yaml, the jsonl_path osatlas_ui_tars_cleaned.jsonl maps to an image folder named osatlas. However I can't find this folder in the repo, nor in aguvis-stage1 and aguvis-stage2.

In addition, filenames in the jsonl file (e.g. windows_20240823_072238_before_screenshot_sub1_6_6.png) don't match the filenames in OS-Copilot/OS-Atlas-data (closest is windows/20240823_072238_before_screenshot.png). What does the _sub1_6_6 mean? And is this data available somewhere?

Thanks a lot

same problem

i also encounter this issue :

DatasetGenerationCastError: An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 1 missing columns ({'id'})

This happened while the json dataset builder was generating data using

hf://datasets/xlangai/Jedi/datasets/icon_caption/training_data_icon/training_data_icon_conversations-images_pure_color_background.jsonl (at revision bad7cb803e535de2b8b6c5c6da92f4539d6e0bcd)

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
XLang NLP Lab org

Oh hi guys, for this part you reused the images from https://huggingface.co/datasets/OS-Copilot/OS-Atlas-data/tree/main/desktop_domain. You downloaded the images from here and aligned the paths with the data items we prepared.

Oh hi guys, for this part you reused the images from https://huggingface.co/datasets/OS-Copilot/OS-Atlas-data/tree/main/desktop_domain. You downloaded the images from here and aligned the paths with the data items we prepared.

I tried, just like @rronan-h mentioned, "In addition, filenames in the jsonl file (e.g. windows_20240823_072238_before_screenshot_sub1_6_6.png) don't match the filenames in OS-Copilot/OS-Atlas-data (closest is windows/20240823_072238_before_screenshot.png). What does the _sub1_6_6 mean? And is this data available somewhere?" how to get the 6_6 suffix?

XLang NLP Lab org

Oh I see. We made internal rename of these images... And I am afraid the data have been cleaned due to company policy. So I recommend if you can matching the data with images by instructios. It would be nice.

Oh I see. We made internal rename of these images... And I am afraid the data have been cleaned due to company policy. So I recommend if you can matching the data with images by instructios. It would be nice.

Could you please provide the renaming and processing detailes for os-atlas datasets, thanks.

XLang NLP Lab org

Lost the code for that part as well... But my conclusion is, the cleaned version of OS-Atlas really really helps a lot. You can even replicate, it is worthy.

Sign up or log in to comment