Add underscores to applychain and extraChain to mark them as internal

Author: Saransh-cpp

src/layers/basic.jl

@@ -48,20 +48,20 @@ end
 
 @functor Chain
 
-(c::Chain)(x) = applychain(c.layers, x)
+(c::Chain)(x) = _applychain(c.layers, x)
 
 # Calculates the forward results of the complete chain provided as a `Tuple`, `AbstractVector`,
 # or a `NamedTuple` of layers with `x` as model input. Users are encouraged to call a chain
 # instead of using this function directly.
-@generated function applychain(layers::Tuple{Vararg{<:Any,N}}, x) where {N}
+@generated function _applychain(layers::Tuple{Vararg{<:Any,N}}, x) where {N}
   symbols = vcat(:x, [gensym() for _ in 1:N])
   calls = [:($(symbols[i+1]) = layers[$i]($(symbols[i]))) for i in 1:N]
   Expr(:block, calls...)
 end
 
-applychain(layers::NamedTuple, x) = applychain(Tuple(layers), x)
+_applychain(layers::NamedTuple, x) = _applychain(Tuple(layers), x)
 
-function applychain(layers::AbstractVector, x)  # type-unstable path, helps compile times
+function _applychain(layers::AbstractVector, x)  # type-unstable path, helps compile times
   for f in layers
     x = f(x)
   end
@@ -102,14 +102,14 @@ julia> activations(c, 1)
 (2, 4, 64)
 ```
 """
-activations(c::Chain, input) = extraChain(Tuple(c.layers), input)
+activations(c::Chain, input) = _extraChain(Tuple(c.layers), input)
 
 # Calculates the forward results of each layer provided in a `Tuple` with `x` as model input.
-function extraChain(fs::Tuple, x)
+function _extraChain(fs::Tuple, x)
   res = first(fs)(x)
-  return (res, extraChain(Base.tail(fs), res)...)
+  return (res, _extraChain(Base.tail(fs), res)...)
 end
-extraChain(::Tuple{}, x) = ()
+_extraChain(::Tuple{}, x) = ()
 
 
 """

src/layers/conv.jl

@@ -76,8 +76,9 @@ Keywords to control initialization of the layer:
 * `bias` - The initial bias vector is all zero by default. Trainable bias can be disabled entirely
   by setting this to `false`, or another vector can be provided such as `bias = randn(Float32, out)`.
 
-Convolutional layers can also be constructed with given weights and biases. This constructor accepts the
-same keywords (and has the same defaults) as the `Conv((4,4), 3 => 7, relu)` method.
+Convolutional layer can also be manually constructed by passing in weights and
+biases. This constructor accepts the same keywords (and has the same defaults)
+as the `Conv((4,4), 3 => 7, relu)` method.
   
 See also [`ConvTranspose`](@ref), [`DepthwiseConv`](@ref), [`CrossCor`](@ref).
 
@@ -212,7 +213,7 @@ Note that `pad=SamePad()` here tries to ensure `size(output,d) == size(x,d) * st
 Parameters are controlled by additional keywords, with defaults
 `init=glorot_uniform` and `bias=true`.
 
-ConvTranspose layer can also be manually constructed with passing in weights and
+ConvTranspose layer can also be manually constructed by passing in weights and
 biases. This constructor accepts the same keywords (and has the same defaults) as the
 `ConvTranspose((4,4), 3 => 7, relu)` method.
 
@@ -351,7 +352,7 @@ specifying the size of the convolutional kernel;
 Parameters are controlled by additional keywords, with defaults
 `init=glorot_uniform` and `bias=true`.
 
-CrossCor layer can also be manually constructed with passing in weights and
+CrossCor layer can also be manually constructed by passing in weights and
 biases. This constructor accepts the layer accepts the same keywords (and has
 the same defaults) as the `CrossCor((4,4), 3 => 7, relu)` method.