new struct syntax

src/backend/mxnet/model.jl

@@ -1,6 +1,6 @@
 using Flux: runrawbatched
 
-type AlterParam
+struct AlterParam
   param
   load
   store
@@ -15,7 +15,7 @@ function copyargs!(as, bs)
   end
 end
 
-type Graph
+struct Graph
   output
   params::Dict{Symbol,Any}
   stacks::Dict{Any,Any}
@@ -31,7 +31,7 @@ end
 
 ndparams(d::Dict{Symbol,MXArray}) = Dict(k => v.data for (k, v) in d)
 
-type Exec
+struct Exec
   graph::Graph
   exec::mx.Executor
   args::Dict{Symbol,MXArray}
@@ -84,7 +84,7 @@ end
 
 # TODO: if `last` changes, update params appropriately
 
-type Model <: Flux.Model
+mutable struct Model <: Flux.Model
   model::Any
   graph::Graph
   execs::Dict{Tuple,Exec}
@@ -119,7 +119,7 @@ Flux.update!(m::Model, η) = (update!(m.last, η); m)
 
 # MX FeedForward interface
 
-type SoftmaxOutput
+struct SoftmaxOutput
   name::Symbol
 end
 

src/backend/mxnet/mxarray.jl

@@ -5,7 +5,7 @@ using MXNet
 
 reversedims!(dest, xs) = permutedims!(dest, xs, ndims(xs):-1:1)
 
-immutable MXArray{N}
+struct MXArray{N}
   data::mx.NDArray
   scratch::Array{Float32,N}
 end

src/backend/tensorflow/model.jl

@@ -1,4 +1,4 @@
-type Model
+struct Model
   model::Any
   session::Session
   params::Dict{Flux.Param,Tensor}

src/backend/tensorflow/recurrent.jl

@@ -1,6 +1,6 @@
 # TODO: refactor, some of this is more general than just the TF backend
 
-type SeqModel
+struct SeqModel
   m::Model
   state::Any
 end

src/backend/tensorflow/tensorflow.jl

@@ -5,7 +5,7 @@ import Flux: accuracy, rebatch, convertel
 
 export tf
 
-type Op
+struct Op
   f
   shape
 end

src/compiler/loops.jl

@@ -1,6 +1,6 @@
 export unroll, unroll1
 
-type Offset
+struct Offset
   name::Symbol
   n::Int
   default::Nullable{Param}

src/compiler/shape.jl

@@ -2,7 +2,7 @@ using DataFlow.Interpreter
 
 export @shapes
 
-type Hint
+struct Hint
   typ
 end
 

src/dims/batching.jl

@@ -1,6 +1,6 @@
 export Batch, batchone
 
-immutable Batch{T,S} <: AbstractVector{T}
+struct Batch{T,S} <: AbstractVector{T}
   data::CatMat{T,S}
 end
 

src/dims/catmat.jl

@@ -2,7 +2,7 @@ import Base: eltype, size, getindex, setindex!, convert
 
 export CatMat, rawbatch
 
-immutable CatMat{T,S} <: AbstractVector{T}
+struct CatMat{T,S} <: AbstractVector{T}
   data::S
 end
 

src/dims/seq.jl

@@ -1,6 +1,6 @@
 export seq, Seq, BatchSeq
 
-immutable Seq{T,S} <: AbstractVector{T}
+struct Seq{T,S} <: AbstractVector{T}
   data::CatMat{T,S}
 end
 

src/layers/shape.jl

@@ -11,12 +11,12 @@ single(i::Dims) = length(i) == 1 ? first(i) : i
 
 # Shim for kicking off shape inference
 
-type ShapeError <: Exception
+struct ShapeError <: Exception
   layer
   shape
 end
 
-type Input{N} <: Model
+struct Input{N} <: Model
   dims::Dims{N}
 end
 
@@ -27,7 +27,7 @@ back!(::Input, Δ, x) = Δ
 
 # Initialise placeholder
 
-type Init{F}
+struct Init{F}
   f::F
 end
 

src/layers/shims.jl

@@ -1,6 +1,6 @@
 export Conv2D, MaxPool, AvgPool, Reshape
 
-type Conv2D <: Model
+struct Conv2D <: Model
   filter::Param{Array{Float64,4}} # [height, width, inchans, outchans]
   stride::Dims{2}
 end
@@ -16,7 +16,7 @@ shape(c::Conv2D, in::Dims{3}) =
 
 for Pool in :[MaxPool, AvgPool].args
   @eval begin
-    type $Pool <: Model
+    struct $Pool <: Model
       size::Dims{2}
       stride::Dims{2}
     end
@@ -34,7 +34,7 @@ for Pool in :[MaxPool, AvgPool].args
   end
 end
 
-immutable Reshape{N}
+struct Reshape{N}
   dims::Dims{N}
 end
 

src/model.jl

@@ -54,7 +54,7 @@ A `Param` object stores a parameter array along with an accumulated delta to
 that array. When converting to backends like TensorFlow, identical `Param`s will
 result in identical variable objects, making model reuse trivial.
 """
-type Param{T}
+struct Param{T}
   x::T
   Δx::T
 end
@@ -107,7 +107,7 @@ Base.copy!(p::Param, xs) = copy!(p.x, xs)
 
 export Capacitor
 
-type Capacitor <: Model
+struct Capacitor <: Model
   graph::IVertex{Any}
 end