Data organization issue #39
Description
Context:
For running a dataset, we need to specify a task_name to pipeline.py, and this task_name will be the directory containing the datasets, for example, /hypothesis-generation/data/task_A
Then when running the task, a BaseTask object will be created using this task_name, and then retrieving the data, metadata, config, etc.
In the config.yaml files, users need to specify another "task_name", which will be only used to find its extract_label register.
The duplicate definition of task_name can cause some confusion or bugs.
Maybe we can change the name or handling of this. We will likely need more organized datasets in the future as the number of datasets is growing quickly, and sometimes one task can have different configurations, so creating multiple extract_label functions for the same task is unnecessary.