This pull request addresses a bug in the multisession training functionality of both the sklearn and torch APIs. Previously, irrespective of the user's preference for GPU execution, the training process was limited to CPU only. Our fix ensures that specified accelerators, such as cuda, are utilized. The problem resided in the DatasetCollection, where CPU was set as the default and users were unable to modify this behavior.
This pull request addresses a bug in the multisession training functionality of both the sklearn and torch APIs. Previously, irrespective of the user's preference for GPU execution, the training process was limited to CPU only. Our fix ensures that specified accelerators, such as cuda, are utilized. The problem resided in the
DatasetCollection, where CPU was set as the default and users were unable to modify this behavior.Fix https://github.com/AdaptiveMotorControlLab/CEBRA-dev/pull/651.