Gated DeltaNet unused D parameter

[koceja]

Hi,

When using the Gated DeltaNet layer, I seem to be running into an issue with reloading the optimizer state from a saved checkpoint. I believe it's because there is an unused parameter 'D' in the layer.

I might be mistaken though, do you know if this parameter is actually unused or maybe its not used with certain configs?

Thanks!

[sustcsonglin]

Just deleted it, thanks for reporting!