fn
batch_norm3d_op
→TensorImplPtrint batch_norm3d_op(const int & x, const int & gamma, const int & beta, double eps, const int & running_mean, const int & running_var, double momentum)Public entry point for 3-D Batch Normalization.
Applies training-mode batch normalization over the batch and three spatial axes — used by 3-D CNNs for volumetric data (video, medical imaging).
See Also
BatchNorm3dBackward : Autograd node implementing the forward + backward.
Parameters
xTensorImplPtrInput tensor of shape
(B, C, D, H, W).gammaTensorImplPtrPer-channel scale of shape
(C,).betaTensorImplPtrPer-channel shift of shape
(C,).epsdouble= NoneNumerical-stability constant. Default:
1e-5.Returns
TensorImplPtrNormalised output, same shape as x.