Shubham Agrawal
Shubham Agrawal
I think there is no explicit method to do so yet. But I think you can use optimizer arguments to train easily.
Is this issue still open or is it done already?
https://github.com/mlpack/mlpack/pull/3231#discussion_r917359361
https://github.com/mlpack/mlpack/pull/3240/commits/36da227d2640e915e8fc92cd4ea0bb13ff1927cf This commit corrects the backward pass for the convolution layer.
@rcurtin You can start reviewing this PR.
> It looks like some of the tests are failing; can you look into them please? > > ``` > 2022-08-20T09:23:13.6520380Z ------------------------------------------------------------------------------- > 2022-08-20T09:23:13.6520840Z AdvancedConvolutionLayerTest > 2022-08-20T09:23:13.6521910Z ------------------------------------------------------------------------------- > 2022-08-20T09:23:13.6522870Z...
@rcurtin You can review this. There are some style issues. Except that I think it is good to merge this.
I have added an additional test for the Convolution layer, which verifies the correctness of forward call and backpropagation. I have tested with PyTorch implementation.
> Nice, this is looking good! I know that the code is very similar to `Convolution`, but I think that it is good to keep the additional complexity of the...
@rcurtin I think you can merge this now.