Neural Nets training with multiple Chains Lux.jl and LuxCUDA.jl: LoadError: AssertionError: length(init_params) == length(depvars) when using multi chains in NeuralPDE.jl GPU example

It seems that I have the same problem as Leo (Neural Nets training with multiple Chains Lux.jl and CUDA.jl), but its solution does not work for me. First, CUDA has been replaced by LuxCUDA in the current version. Second, even after I changed to LuxCUDA and changed ‘gpu’ to ‘gpud’, the scalar error is still reported.