infill_loss Namespace Reference

Classes
class	InfillLossCollection
class	InfillLossInduction

Functions
def	_assert_no_grad (variable)

Detailed Description

HolePixelLoss
This loss mimics nividia's pixelwise loss for holes (L1)
used in the infill network

Loss function designed for a network that infills tracks. Separate functions for collection views
(unipolar signals) and induction views (bipolar signal)
Taken from github.com/NuTufts/sparse_infill.

Function Documentation

def infill_loss._assert_no_grad (   variable )

private

Definition at line 16 of file infill_loss.py.

def _assert_no_grad(variable):
    assert not variable.requires_grad, \
        "nn criterions don't compute the gradient w.r.t. targets - please " \
        "mark these variables as not requiring gradients"