We would specify a small number of models that are pretrained on common use cases for the users’ convenience, these are small enough s.t. they can be provided via Github directly.
This could be achieved by using autoML (hyperparameter tuning), tie in with slurm, an experiment tracker (Neptune.ai currently, mlflow as a goal). We can think about serving the models. Ties well into Faruk Diblen machine learning on Snellius pipeline / workflow.
We would specify a small number of models that are pretrained on common use cases for the users’ convenience, these are small enough s.t. they can be provided via Github directly.
This could be achieved by using autoML (hyperparameter tuning), tie in with slurm, an experiment tracker (Neptune.ai currently, mlflow as a goal). We can think about serving the models. Ties well into Faruk Diblen machine learning on Snellius pipeline / workflow.