Hi,
I understand that you are encountering an error related to "cuDNN" when performing convolution operations. This issue can occur during training or inference (using "predict", "classify", or "activations") with a convolutional "dlnetwork" or "DAGNetwork". The error might arise when networks with particular input sizes perform convolution operations or when a convolution layer is followed by a ReLU layer.
This is a known bug in R2023a and has been reported in the following bug report:
To fix this, I would suggest updating to R2023a Update 3 or upgrading to R2023b, as this resolves the issue.
Hope this helps!
