Method error when using Flux.setup with Embedding layer

[egolep]

Whenever I try to use the Flux.setup function, if my neural network (Chain) contains an Embedding layer, I get this error:

ERROR: MethodError: no method matching Embedding(::Bool)
Closest candidates are:
  Embedding(::Integer, ::Integer; kw...) at ~/.julia/packages/Flux/kq9Et/src/deprecations.jl:67
  Embedding(::Pair{<:Integer, <:Integer}; init) at ~/.julia/packages/Flux/kq9Et/src/layers/basic.jl:694
  Embedding(::W) where W<:(AbstractMatrix) at ~/.julia/packages/Flux/kq9Et/src/layers/basic.jl:689

For example, this snippet fails:

m = Embedding(3=>1)
Flux.setup(Adam(), m)

It fails both on CPU and GPU

[mcabbott]

That's no good, sorry.

Seems like a bug in Flux, since the supposedly equivalent Optimisers.jl functions work fine -- and can be used as a work-around for now:

julia> Optimisers.setup(Optimisers.Adam(), Embedding(3 => 1))
(weight = Leaf(Adam{Float32}(0.001, (0.9, 0.999), 1.19209f-7), (Float32[0.0 0.0 0.0], Float32[0.0 0.0 0.0], (0.9, 0.999))),)

julia> Flux.setup(Flux.Adam(), Embedding(3=>1))
ERROR: MethodError: no method matching Embedding(::Bool)
...
Stacktrace:
  [1] (::Flux.var"#222#223")(y::NamedTuple{(:weight,), Tuple{Bool}})
    @ Flux ~/.julia/packages/Functors/orBYx/src/functor.jl:20
  [2] (::Functors.DefaultWalk)(::Function, ::Embedding{Matrix{Float32}})
    @ Functors ~/.julia/packages/Functors/orBYx/src/walks.jl:62
  [3] ExcludeWalk
    @ ~/.julia/packages/Functors/orBYx/src/walks.jl:92 [inlined]
  [4] CachedWalk
    @ ~/.julia/packages/Functors/orBYx/src/walks.jl:132 [inlined]
  [5] fmap
    @ ~/.julia/packages/Functors/orBYx/src/maps.jl:1 [inlined]
  [6] #fmap#25
    @ ~/.julia/packages/Functors/orBYx/src/maps.jl:11 [inlined]
  [7] setup(rule::Optimisers.Adam{Float64}, model::Embedding{Matrix{Float32}})
    @ Flux.Train ~/.julia/packages/Flux/kq9Et/src/train.jl:51
  [8] setup(rule::Adam, model::Embedding{Matrix{Float32}})
    @ Flux ~/.julia/packages/Flux/kq9Et/src/deprecations.jl:117

[egolep]

Nothing to be sorry about.
Thanks for the quick reply and fix!