tests

Project.toml

@@ -26,14 +26,4 @@ ConstructionBase = "1.5.8"
 EnzymeCore = "0.8.5"
 Functors = "0.4.9, 0.5"
 Statistics = "1"
-Zygote = "0.6.40, 0.7.1"
 julia = "1.10"
-
-[extras]
-EnzymeCore = "f151be2c-9106-41f4-ab19-57ee4f262869"
-StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
-Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
-Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
-
-[targets]
-test = ["Test", "EnzymeCore", "StaticArrays", "Zygote"]

src/interface.jl

@@ -273,7 +273,7 @@ macro def(expr)
   # Positional-argument method, has defaults for all but the first arg:
   positional = :(function $rule($(names[1]), $(params[2:end]...))
     $check_sign_eta
-    vars = maybe_float.([$(names...)])
+    vars = $(maybe_float).(($(names...)),($(default_types...)))
     return new{typeof.(vars)...}(vars...)
   end)
   # Keyword-argument method. (Made an inner constructor only to allow
@@ -283,5 +283,5 @@ macro def(expr)
   return esc(expr)
 end
 
-maybe_float(x::Number) = float(x)
-maybe_float(x) = x
+maybe_float(x, T::Type{<:AbstractFloat}) = float(x)
+maybe_float(x, T) = x

src/rules.jl

@@ -114,16 +114,16 @@ gradients by an estimate their variance, instead of their second moment.
 - Keyword `centred` (or `centered`): Indicates whether to use centred variant
                                      of the algorithm.
 """
-struct RMSProp <: AbstractRule
-  eta::Float64
-  rho::Float64
-  epsilon::Float64
+struct RMSProp{Teta,Trho,Teps} <: AbstractRule
+  eta::Teta
+  rho::Trho
+  epsilon::Teps
   centred::Bool
 end
 
 function RMSProp(η, ρ = 0.9, ϵ = 1e-8; centred::Bool = false, centered::Bool = false)
   η < 0 && throw(DomainError(η, "the learning rate cannot be negative"))
-  RMSProp(η, ρ, ϵ, centred | centered)
+  return RMSProp(float(η), float(ρ), float(ϵ), centred | centered)
 end
 RMSProp(; eta = 0.001, rho = 0.9, epsilon = 1e-8, kw...) = RMSProp(eta, rho, epsilon; kw...)
 
@@ -155,7 +155,7 @@ end
 
 
 """
-    Rprop(η = 1f-3, ℓ = (5f-1, 1.2f0), Γ = (1f-6, 50f0))
+    Rprop(η = 1e-3, ℓ = (0.5, 1.2), Γ = (1e-6, 50.0))
     Rprop(; [eta, ell, gamma])
 
 Optimizer using the
@@ -171,9 +171,9 @@ learning algorithm that depends only on the sign of the gradient.
 - Step sizes (`Γ::Tuple == gamma`): Mminimal and maximal allowed step sizes.
 """
 @def struct Rprop <: AbstractRule
-    eta =  1f-3
-    ell = (5f-1, 1.2f0)
-    gamma = (1f-6, 50f0)
+    eta =  1e-3
+    ell = (0.5, 1.2)
+    gamma = (1e-6, 50.0)
 end
 
 init(o::Rprop, x::AbstractArray) = (zero(x), onevalue(o.eta, x))
@@ -528,17 +528,17 @@ Implemented as an [`OptimiserChain`](@ref) of [`Adam`](@ref) and [`WeightDecay`]
     The previous rule, which is closer to the original paper, can be obtained by setting `AdamW(..., couple=false)`.
     See [this issue](https://github.com/FluxML/Flux.jl/issues/2433) for more details.
 """
-struct AdamW <: AbstractRule
-  eta::Float64
-  beta::Tuple{Float64, Float64}
-  lambda::Float64
-  epsilon::Float64
+struct AdamW{Teta,Tbeta<:Tuple,Tlambda,Teps} <: AbstractRule
+  eta::Teta
+  beta::Tbeta
+  lambda::Tlambda
+  epsilon::Teps
   couple::Bool
 end
 
 function AdamW(η, β = (0.9, 0.999), λ = 0.0, ϵ = 1e-8; couple::Bool = true)
   η < 0 && throw(DomainError(η, "the learning rate cannot be negative"))
-  AdamW(η, β, λ, ϵ, couple)
+  return AdamW(float(η), β, float(λ), float(ϵ), couple)
 end
 
 AdamW(; eta = 0.001, beta = (0.9, 0.999), lambda= 0.0,  epsilon = 1e-8, kw...) =
@@ -704,12 +704,12 @@ Typically composed with other rules using [`OptimiserChain`](@ref).
 
 See also [`ClipGrad`](@ref).
 """
-struct ClipNorm <: AbstractRule
-  omega::Float64
-  p::Float64
+struct ClipNorm{To,Tp} <: AbstractRule
+  omega::To
+  p::Tp
   throw::Bool
 end
-ClipNorm(ω = 10, p = 2; throw::Bool = true) = ClipNorm(ω, p, throw)
+ClipNorm(ω = 10, p = 2; throw::Bool = true) = ClipNorm(float(ω), float(p), throw)
 
 init(o::ClipNorm, x::AbstractArray) = nothing
 

test/Project.toml

@@ -0,0 +1,8 @@
+[deps]
+ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
+EnzymeCore = "f151be2c-9106-41f4-ab19-57ee4f262869"
+Functors = "d9f16b24-f501-4c13-a1f2-28368ffc5196"
+Optimisers = "3bd65402-5787-11e9-1adc-39752487f4e2"
+StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
+Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"

test/interface.jl

@@ -0,0 +1,24 @@
+@testset "@def" begin 
+    Optimisers.@def struct DummyRule
+        a = 1
+        b1 = 1.5
+        b2 = 2.5f0
+        c = (1.0, 2.0)
+    end
+
+    # no args
+    r = DummyRule()
+    @test typeof(r.a) == Int
+    @test typeof(r.b1) == Float64
+    @test typeof(r.b2) == Float32
+    @test typeof(r.c) == Tuple{Float64, Float64}
+
+    # some positional args    
+    r = DummyRule(2, 2, 4.5)
+    @test r.a == 2
+    @test r.b1 == 2
+    @test r.b2 == 4.5
+    @test typeof(r.b1) == Float64 # int promoted to float
+    @test typeof(r.b2) == Float64 # Float64 not converted to Float32
+    @test r.c == (1.0, 2.0)
+end

test/runtests.jl

@@ -557,4 +557,7 @@ end
   @testset verbose=true "Optimisation Rules" begin
     include("rules.jl")
   end
+  @testset verbose=true "interface" begin
+    include("interface.jl")
+  end
 end