Skip to content

the derivative for 'cudnn_batch_norm_backward save_mean' is not implemented #4422

@elbaro

Description

@elbaro
loss_gp = ..
loss_gp.backward()  # error

Having this error when BatchNorm + wgan_gp (second order) is used.
It's installed from master fec3d4.

Metadata

Metadata

Assignees

Labels

in progressmodule: cudnnRelated to torch.backends.cudnn, and CuDNN supportmodule: double backwardsProblem is related to double backwards definition on an operator

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions