You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello, thank you very much for your work in implementing KAN. I am currently looking to integrate the KANs you developed into ModelingToolkit.jl, which requires the representation of equations. The approach I am following is based on ModelingToolkitNeuralNets.jl, where the core idea is to convert the LuxCore.stateless_apply method into equations.
However, after trying this approach, I found that the current version of KolmogorovArnold.jl does not support LuxCore.Internal.get_empty_state, which prevents me from using LuxCore.stateless_apply to convert it into equations. To make this work, and in alignment with the model algorithms, I would like to request that the repository add an interface for LuxCore.Internal.get_empty_state, as follows:
Hello, thank you very much for your work in implementing KAN. I am currently looking to integrate the KANs you developed into ModelingToolkit.jl, which requires the representation of equations. The approach I am following is based on ModelingToolkitNeuralNets.jl, where the core idea is to convert the
LuxCore.stateless_apply
method into equations.However, after trying this approach, I found that the current version of KolmogorovArnold.jl does not support
LuxCore.Internal.get_empty_state
, which prevents me from usingLuxCore.stateless_apply
to convert it into equations. To make this work, and in alignment with the model algorithms, I would like to request that the repository add an interface forLuxCore.Internal.get_empty_state
, as follows:I am not sure if this method is correct, but I hope you can add support for this method in your future work. Thank you!
The text was updated successfully, but these errors were encountered: