Description
In a DAG we can currently specify different params and their default value however I'd like to have a list of presets from which I can load on the fly different pre-configured default values and load them on the fly and eventually edit the preset's default values from the UI if needed.
Use case/motivation
I often need to to run Machine learning training experiments with the same model architecture but different dataset for different tasks, since what changes is the dataset but not the architectures I use a dag to fetch data, train, evaluated etc. However different datasets need different augmentations techniques which often have lots of parameters and having multiple presets could help engineers start the training with a preselected config and maybe tweak only few parameters rather than rewriting the whole augmentations
Related issues
No response
Are you willing to submit a PR?
Code of Conduct
Description
In a DAG we can currently specify different params and their default value however I'd like to have a list of presets from which I can load on the fly different pre-configured default values and load them on the fly and eventually edit the preset's default values from the UI if needed.
Use case/motivation
I often need to to run Machine learning training experiments with the same model architecture but different dataset for different tasks, since what changes is the dataset but not the architectures I use a dag to fetch data, train, evaluated etc. However different datasets need different augmentations techniques which often have lots of parameters and having multiple presets could help engineers start the training with a preselected config and maybe tweak only few parameters rather than rewriting the whole augmentations
Related issues
No response
Are you willing to submit a PR?
Code of Conduct