Closed AviAvni closed 6 years ago
What do you think of using the existing Stride
type which is already used in the Conv2D.pooling
layer?
https://github.com/mathias-brandewinder/CNTK.FSharp/blob/31ba3f0346fcf08bcc5f0d6c27b1795758a82c41/CNTK.FSharp/Sequential.fs#L138-L141
As far as I know, in the case of a 2D convolution, there should be 2 arguments for the stride, so it seems to me that enforcing it by using a record would be slightly better.
@mathias-brandewinder you right thanks
Thank you 👍
See https://github.com/mathias-brandewinder/CNTK.FSharp/issues/12