Action Normalization of gripper in bridge dataset between octo-small and ovto-small-1.5

[cccedric]

Hi author!

Thanks for your contribution to this amazing model!

When I test the model, I found that the data statistic for bridge dataset is different between octo-small and octo-small-1.5. Specifically for bridge dataset, the [min, max, mean] of gripper action for octo-small is [-1, 1, 0] and for octo-small-1.5 is [0, 1, 0.58]. Did you change the transform function when training these 2 models?

Thanks!

Best,
cedric