-
Notifications
You must be signed in to change notification settings - Fork 168
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
How to adapt the models to grid data and the correspondence between data and models? #368
Comments
Here is how to generalize models used for point-based data for grid-based data. (1) If the dataset class used by the model is TrafficStatePointDataset, such as AGCRN, ASTGCNCommon, CCRNN, etc., you can directly set dataset_class to TrafficStateGridDataset in task_file.json or through a custom configuration file(--config_file). Then set the parameter use_row_column of TrafficStateGridDataset to False. (2) If the dataset class used by the model is the subclass of TrafficStatePointDataset, such as ASTGCNDataset, CONVGCNDataset, STG2SeqDataset, etc., you can modify the file of the dataset class to make it inherit TrafficStateGridDataset instead of the current TrafficStatePointDataset. Then set the parameter use_row_column in the function init() to False. Example (1):Before modification:task_config.json After modification:task_config.json Example (2)::Before modification:task_config.json After modification:task_config.json |
Additional processing may be required for some special models. |
For the details, please visit https://bigscity-libcity-docs.readthedocs.io/en/latest/user_guide/data/dataset_for_task.html
Note 5. There are some models that require three inputs, i.e., CLOSENESS, PERIOD, TREND, such as STResNet, ACFM, ASTGCN. for such models we implemented corresponding generalized versions with only CLOSENESS inputs for fair comparisons, i.e., STResNetCommon, ACFMCommon, ASTGCNCommon.
The text was updated successfully, but these errors were encountered: