[GNNLux] updates for Lux v1.0

GNNLux/Project.toml

@@ -12,14 +12,16 @@ LuxCore = "bb33d45b-7691-41d6-9220-0943567d0623"
 NNlib = "872c559c-99b0-510c-b3b7-b6c96a88d5cd"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
+Static = "aedffcd0-7271-4cad-89d0-dc628f76c6d3"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 
 [compat]
 ConcreteStructs = "0.2.3"
-Lux = "0.5.61"
-LuxCore = "0.1.20"
+Lux = "1.0"
+LuxCore = "1.0"
 NNlib = "0.9.21"
 Reexport = "1.2"
+Static = "1.1"
 julia = "1.10"
 
 [extras]

GNNLux/src/GNNLux.jl

@@ -2,14 +2,15 @@ module GNNLux
 using ConcreteStructs: @concrete
 using NNlib: NNlib, sigmoid, relu, swish
 using Statistics: mean
-using LuxCore: LuxCore, AbstractExplicitLayer, AbstractExplicitContainerLayer, parameterlength, statelength, outputsize, 
+using LuxCore: LuxCore, AbstractLuxLayer, AbstractLuxContainerLayer, parameterlength, statelength, outputsize, 
               initialparameters, initialstates, parameterlength, statelength
 using Lux: Lux, Chain, Dense, GRUCell,
            glorot_uniform, zeros32, 
            StatefulLuxLayer
 using Reexport: @reexport
 using Random: AbstractRNG
 using GNNlib: GNNlib
+using Static
 @reexport using GNNGraphs
 
 include("layers/basic.jl")

GNNLux/src/layers/basic.jl

@@ -1,14 +1,14 @@
 """
-    abstract type GNNLayer <: AbstractExplicitLayer end
+    abstract type GNNLayer <: AbstractLuxLayer end
 
 An abstract type from which graph neural network layers are derived.
-It is Derived from Lux's `AbstractExplicitLayer` type.
+It is Derived from Lux's `AbstractLuxLayer` type.
 
 See also [`GNNChain`](@ref GNNLux.GNNChain).
 """
-abstract type GNNLayer <: AbstractExplicitLayer end
+abstract type GNNLayer <: AbstractLuxLayer end
 
-abstract type GNNContainerLayer{T} <: AbstractExplicitContainerLayer{T} end
+abstract type GNNContainerLayer{T} <: AbstractLuxContainerLayer{T} end
 
 @concrete struct GNNChain <: GNNContainerLayer{(:layers,)}
     layers <: NamedTuple
@@ -24,7 +24,7 @@ function GNNChain(; kw...)
     return GNNChain(nt)
 end
 
-_wrapforchain(l::AbstractExplicitLayer) = l
+_wrapforchain(l::AbstractLuxLayer) = l
 _wrapforchain(l) = Lux.WrappedFunction(l)
 
 Base.keys(c::GNNChain) = Base.keys(getfield(c, :layers))
@@ -44,7 +44,7 @@ Base.firstindex(c::GNNChain) = firstindex(c.layers)
 
 LuxCore.outputsize(c::GNNChain) = LuxCore.outputsize(c.layers[end])
 
-(c::GNNChain)(g::GNNGraph, x, ps, st) = _applychain(c.layers, g, x, ps, st)
+(c::GNNChain)(g::GNNGraph, x, ps, st) = _applychain(c.layers, g, x, ps.layers, st.layers)
 
 function _applychain(layers, g::GNNGraph, x, ps, st)  # type-unstable path, helps compile times
     newst = (;)
@@ -56,6 +56,6 @@ function _applychain(layers, g::GNNGraph, x, ps, st)  # type-unstable path, help
 end
 
 _applylayer(l, g::GNNGraph, x, ps, st) = l(x), (;)
-_applylayer(l::AbstractExplicitLayer, g::GNNGraph, x, ps, st) = l(x, ps, st)
+_applylayer(l::AbstractLuxLayer, g::GNNGraph, x, ps, st) = l(x, ps, st)
 _applylayer(l::GNNLayer, g::GNNGraph, x, ps, st) = l(g, x, ps, st)
 _applylayer(l::GNNContainerLayer, g::GNNGraph, x, ps, st) = l(g, x, ps, st)

GNNLux/src/layers/conv.jl

@@ -1,7 +1,9 @@
 _getbias(ps) = hasproperty(ps, :bias) ? getproperty(ps, :bias) : false
 _getstate(st, name) = hasproperty(st, name) ? getproperty(st, name) : NamedTuple()
 _getstate(s::StatefulLuxLayer{true}) = s.st
+_getstate(s::StatefulLuxLayer{Static.True}) = s.st
 _getstate(s::StatefulLuxLayer{false}) = s.st_any
+_getstate(s::StatefulLuxLayer{Static.False}) = s.st_any
 
 
 @concrete struct GCNConv <: GNNLayer
@@ -20,10 +22,9 @@ function GCNConv(ch::Pair{Int, Int}, σ = identity;
                 init_bias = zeros32,
                 use_bias::Bool = true,
                 add_self_loops::Bool = true,
-                use_edge_weight::Bool = false,
-                allow_fast_activation::Bool = true)
+                use_edge_weight::Bool = false)
     in_dims, out_dims = ch
-    σ = allow_fast_activation ? NNlib.fast_act(σ) : σ
+    σ = NNlib.fast_act(σ)
     return GCNConv(in_dims, out_dims, use_bias, add_self_loops, use_edge_weight, init_weight, init_bias, σ)
 end
 
@@ -121,10 +122,9 @@ function GraphConv(ch::Pair{Int, Int}, σ = identity;
             aggr = +,
             init_weight = glorot_uniform,
             init_bias = zeros32, 
-            use_bias::Bool = true, 
-            allow_fast_activation::Bool = true)
+            use_bias::Bool = true)
     in_dims, out_dims = ch
-    σ = allow_fast_activation ? NNlib.fast_act(σ) : σ
+    σ = NNlib.fast_act(σ)
     return GraphConv(in_dims, out_dims, use_bias, init_weight, init_bias, σ, aggr)
 end
 
@@ -212,11 +212,10 @@ end
 CGConv(ch::Pair{Int, Int}, args...; kws...) = CGConv((ch[1], 0) => ch[2], args...; kws...)
 
 function CGConv(ch::Pair{NTuple{2, Int}, Int}, act = identity; residual = false,
-                use_bias = true, init_weight = glorot_uniform, init_bias = zeros32, 
-                allow_fast_activation = true)
+                use_bias = true, init_weight = glorot_uniform, init_bias = zeros32)
     (nin, ein), out = ch
-    dense_f = Dense(2nin + ein => out, sigmoid; use_bias, init_weight, init_bias, allow_fast_activation)
-    dense_s = Dense(2nin + ein => out, act; use_bias, init_weight, init_bias, allow_fast_activation)
+    dense_f = Dense(2nin + ein => out, sigmoid; use_bias, init_weight, init_bias)
+    dense_s = Dense(2nin + ein => out, act; use_bias, init_weight, init_bias)
     return CGConv((nin, ein), out, dense_f, dense_s, residual, init_weight, init_bias)
 end
 
@@ -232,7 +231,7 @@ function (l::CGConv)(g, x, e, ps, st)
 end
 
 @concrete struct EdgeConv <: GNNContainerLayer{(:nn,)}
-    nn <: AbstractExplicitLayer
+    nn <: AbstractLuxLayer
     aggr
 end
 
@@ -246,10 +245,10 @@ end
 
 
 function (l::EdgeConv)(g::AbstractGNNGraph, x, ps, st)
-    nn = StatefulLuxLayer{true}(l.nn, ps, st)
+    nn = StatefulLuxLayer{true}(l.nn, ps.nn, st.nn)
     m = (; nn, l.aggr)
     y = GNNlib.edge_conv(m, g, x)
-    stnew = _getstate(nn)
+    stnew = (; nn = _getstate(nn)) # TODO: support also aggr state if present
     return y, stnew
 end
 
@@ -608,18 +607,18 @@ function Base.show(io::IO, l::GatedGraphConv)
 end
 
 @concrete struct GINConv <: GNNContainerLayer{(:nn,)}
-    nn <: AbstractExplicitLayer
+    nn <: AbstractLuxLayer
     ϵ <: Real
     aggr
 end
 
 GINConv(nn, ϵ; aggr = +) = GINConv(nn, ϵ, aggr)
 
 function (l::GINConv)(g, x, ps, st)
-    nn = StatefulLuxLayer{true}(l.nn, ps, st)
+    nn = StatefulLuxLayer{true}(l.nn, ps.nn, st.nn)
     m = (; nn, l.ϵ, l.aggr)
     y = GNNlib.gin_conv(m, g, x)
-    stnew = _getstate(nn)
+    stnew = (; nn = _getstate(nn))
     return y, stnew
 end
 
@@ -669,4 +668,4 @@ function Base.show(io::IO, l::MEGNetConv)
     nout = l.out_dims
     print(io, "MEGNetConv(", nin, " => ", nout)
     print(io, ")")
-end
+end

GNNLux/src/layers/temporalconv.jl

@@ -1,4 +1,4 @@
-@concrete struct StatefulRecurrentCell <: AbstractExplicitContainerLayer{(:cell,)}     
+@concrete struct StatefulRecurrentCell <: AbstractLuxContainerLayer{(:cell,)}     
     cell <: Union{<:Lux.AbstractRecurrentCell, <:GNNContainerLayer}
 end
 
@@ -7,16 +7,16 @@ function LuxCore.initialstates(rng::AbstractRNG, r::GNNLux.StatefulRecurrentCell
 end
 
 function (r::StatefulRecurrentCell)(g, x::AbstractMatrix, ps, st::NamedTuple)
-    (out, carry), st = applyrecurrentcell(r.cell, g, x, ps, st.cell, st.carry)
+    (out, carry), st = applyrecurrentcell(r.cell, g, x, ps.cell, st.cell, st.carry)
     return out, (; cell=st, carry)
 end
 
 function (r::StatefulRecurrentCell)(g, x::AbstractVector, ps, st::NamedTuple)
-    st, carry = st.cell, st.carry
+    stcell, carry = st.cell, st.carry
     for xᵢ in x
-        (out, carry), st = applyrecurrentcell(r.cell, g, xᵢ, ps, st, carry)
+        (out, carry), stcell = applyrecurrentcell(r.cell, g, xᵢ, ps.cell, stcell, carry)
     end
-    return out, (; cell=st, carry)
+    return out, (; cell=stcell, carry)
 end
 
 function applyrecurrentcell(l, g, x, ps, st, carry)
@@ -35,7 +35,7 @@ end
 
 function TGCNCell(ch::Pair{Int, Int}; use_bias = true, init_weight = glorot_uniform, init_state = zeros32, init_bias = zeros32, add_self_loops = false, use_edge_weight = true) 
     in_dims, out_dims = ch
-    conv = GCNConv(ch, sigmoid; init_weight, init_bias, use_bias, add_self_loops, use_edge_weight, allow_fast_activation= true)
+    conv = GCNConv(ch, sigmoid; init_weight, init_bias, use_bias, add_self_loops, use_edge_weight)
     gru = Lux.GRUCell(out_dims => out_dims; use_bias, init_weight = (init_weight, init_weight, init_weight), init_bias = (init_bias, init_bias, init_bias), init_state = init_state)
     return TGCNCell(in_dims, out_dims, conv, gru, init_state)
 end

GNNLux/test/layers/basic_tests.jl

@@ -4,16 +4,16 @@
     x = randn(rng, Float32, 3, 10)        
 
     @testset "GNNLayer" begin
-        @test GNNLayer <: LuxCore.AbstractExplicitLayer
+        @test GNNLayer <: LuxCore.AbstractLuxLayer
     end
 
     @testset "GNNContainerLayer" begin
-        @test GNNContainerLayer <: LuxCore.AbstractExplicitContainerLayer
+        @test GNNContainerLayer <: LuxCore.AbstractLuxContainerLayer
     end
 
     @testset "GNNChain" begin
-        @test GNNChain <: LuxCore.AbstractExplicitContainerLayer{(:layers,)}
-        c = GNNChain(GraphConv(3 => 5, relu), GCNConv(5 => 3))
+        @test GNNChain <: LuxCore.AbstractLuxContainerLayer{(:layers,)}
+        c = GNNChain(GraphConv(3 => 5, tanh), GCNConv(5 => 3))
         test_lux_layer(rng, c, g, x, outputsize=(3,), container=true)
     end
 end

GNNLux/test/layers/conv_tests.jl

@@ -89,7 +89,7 @@
     end
 
     @testset "GINConv" begin
-        nn = Chain(Dense(in_dims => out_dims, relu), Dense(out_dims => out_dims))
+        nn = Chain(Dense(in_dims => out_dims, tanh), Dense(out_dims => out_dims))
         l = GINConv(nn, 0.5)
         test_lux_layer(rng, l, g, x, sizey=(out_dims,g.num_nodes), container=true)
     end