Closed
Milestone
SP15-Item09 - Provide a way to Check the `training_args` Parameter in The Experiment
Milestone ID: 2835
For now the training_args
dictionary os provided by the researcher, propagated to the nodes (via json serializer) and passed as argument to the training_routine (using the **kwargs syntax/hack)
No validity check is made on this dictionary.
We should validate these data:
- presence/absence of key
- range/validity of values (if possible, which may not always be the case)
The validation may depend on the training_plan (torchnn training plan may have different arributes than sklearn training plan)