Parameter data_divide_ratio in input.yaml

[AugustinLu]

Hello,

When I was using SevenNet v0.9.3, I used to set parameter data_divide_ratio in the yaml file to split the original dataset into training and validation sets before training a model.
Example: data_divide_ratio: 0.2

However, after updating to v0.10.2 recently, it seems that this variable seems to no longer have any effect, and a message reporting the absence of validset appears:
Failed to get error recorder key: TotalLoss or validset is missing. There will be no best checkpoint.

Is the parameter data_divide_ratio still in use?

Thank you very much.

[YutackPark]

Sorry for late answer. It is temporarily not working in v0.10.x. I'll make it back as soon as possible. If you want previous training mode with that working that key, you may able to use sevenn -m train_v1 for it. As I didn't tested routine for this legacy mode, use with care.

[AugustinLu]

Thank you for your kind reply.

I understand the situation. For now, we will try either with train_v1, or prepare the training, validation and test sets beforehand.

[YutackPark]

The feature is included in official v0.10.2. Check when you have a time!

[AugustinLu]

I have confirmed that it is working with the latest revision.

Thank you for your swift actions!