`EvalBatchPredict`#

class capymoa.ocl.evaluation.events.EvalBatchPredict[source]#

Bases: TestTaskBegin

After predicting on an evaluation batch, but before any updates.

__init__( train_task: int, test_task: int, continual_eval: int, global_step: int, batch: int, x: Tensor, y: Tensor, y_hat: Tensor, ) → None#

batch: int#: Batch ID within the current task stream.

continual_eval: int#: If multiple evaluations are performed during each training task, this counts which evaluation pass is being performed. Otherwise will be 0.

global_step: int#: A monotonically increasing integer that counts training and evaluation steps.

test_task: int#: The ID of the test task that has begun.

train_task: int#: The ID of the training task that is being evaluated.

x: Tensor#: The input batch.

y: Tensor#: The target batch.

y_hat: Tensor#: The predicted batch.