search

niuzaisheng / ScreenAgent

ScreenAgent: A Computer Control Agent Driven by Visual Language Large Model (IJCAI-24)
https://arxiv.org/abs/2402.07945
Other
217 stars 24 forks source link

Version of sat lib #11

Open HeimingX opened 4 months ago

HeimingX commented 4 months ago

Hi, thanks for the interesting work and the awesome open-sourced codebase.

I'd like to know what version of sat lib you used for ScreenAgent training. In the latest version(0.4.11), the official tutorial for dataset creation is as follows:

def create_dataset_function(path, args):
    # Here to load the dataset
    # ...
    assert isinstance(dataset, torch.utils.data.Dataset)
    return dataset

where create_dataset_function get two parameters, i.e., path and args.

However, in the finetune_ScreenAgent.py, the data creation function does not have those parameters: https://github.com/niuzaisheng/ScreenAgent/blob/85881785f9007d7d336715a593e6f2470c82b4e5/train/finetune_ScreenAgent.py#L284

https://github.com/niuzaisheng/ScreenAgent/blob/85881785f9007d7d336715a593e6f2470c82b4e5/train/dataset/mixture_dataset.py#L37

Could you please help me to solve this problem? Many thanks