Skip to content

Latest commit

 

History

History
23 lines (15 loc) · 1.14 KB

README.md

File metadata and controls

23 lines (15 loc) · 1.14 KB

Datasets preparation

Clone from huggingface

Three required datasets should be cloned from three huggingface datasets.

Using the following commands to download them to this folder

git clone https://huggingface.co/datasets/orlando23/screendata
git clone https://huggingface.co/datasets/orlando23/mobile_pc_web_osworld
git clone https://huggingface.co/datasets/orlando23/failed_agent_trajectory

Double-check the data tracked by lfs downloaded after git clone. Otherwise, use git lfs pull to guarantee they have been synced.

Training and Validation split for ASHL

We provide our used Training and Validation split for ASHL in ScreenReaderData_train.json and ScreenReaderData_val.json. You can take it as your default setting in DINO training to reproduce our result.