Public Member Functions | |
def | __init__ (self, stepsize, threshold, device_option=None, workspace_name="gradient_check", input_device_options=None) |
def | GetLossAndGrad (self, op, grad_ops, inputs, input_names, input_to_check, grad_name, outputs_with_grads) |
def | CheckSimple (self, op, inputs, input_to_check, outputs_with_grads, grad_ops=None, input_device_options=None) |
A gradient checker in Python. This is not the most efficient way to check gradients, as the Python interface will involve a lot of copies back and forth operations. Use at your own risk.
Definition at line 155 of file gradient_checker.py.
def caffe2.python.gradient_checker.GradientChecker.CheckSimple | ( | self, | |
op, | |||
inputs, | |||
input_to_check, | |||
outputs_with_grads, | |||
grad_ops = None , |
|||
input_device_options = None |
|||
) |
Checks the operator in a very simple fashion by stacking a sum of squares on the top. Inputs: op: the operator to be checked. inputs: the input data in numpy arrays. input_to_check: an index specifying which input blob we should check. outputs_with_grads: indices specifying which output blobs will we need to check gradients with. For these outputs, we will collect a squared sum and also feed in their gradients. grad_operator: the gradient operator. If not given, we will get the gradient operator from the gradient registry. input_device_options: an optional mapping from input names to DeviceOptions (to override the default DeviceOption) Outputs: boolean: True if it passes, False if it does not pass.
Definition at line 237 of file gradient_checker.py.