Small fixes

Author: VaclavMacha

docs/src/lecture_11/data/mnist.jl

@@ -11,7 +11,7 @@ function reshape_data(X::AbstractArray{T, 3}, y::AbstractVector) where T
 end
 
 function train_or_load!(file_name, m, X, y; force=false, kwargs...)
-    
+
     !isdir(dirname(file_name)) && mkpath(dirname(file_name))
 
     if force || !isfile(file_name)
@@ -24,8 +24,8 @@ end
 
 function load_data(dataset; onehot=false, T=Float32)
     classes = 0:9
-    X_train, y_train = reshape_data(dataset.traindata(T)...)
-    X_test, y_test = reshape_data(dataset.testdata(T)...)
+    X_train, y_train = reshape_data(dataset(T, :train)[:]...)
+    X_test, y_test = reshape_data(dataset(T, :test)[:]...)
     y_train = T.(y_train)
     y_test = T.(y_test)
 

docs/src/lecture_11/data/mnist_gpu.jl

@@ -38,8 +38,8 @@ end
 
 function load_data(dataset; onehot=false, T=Float32)
     classes = 0:9
-    X_train, y_train = reshape_data(dataset.traindata(T)...)
-    X_test, y_test = reshape_data(dataset.testdata(T)...)
+    X_train, y_train = reshape_data(dataset(T, :train)[:]...)
+    X_test, y_test = reshape_data(dataset(T, :test)[:]...)
     y_train = T.(y_train)
     y_test = T.(y_test)
 

docs/src/lecture_11/exercises.md

@@ -4,11 +4,11 @@ using Flux
 using MLDatasets
 using DataFrames
 using Plots
-using Flux: onehotbatch, onecold, flatten
+using Flux: onehotbatch, onecold, flatten, params
 
 Core.eval(Main, :(using Flux)) # hide
 ENV["DATADEPS_ALWAYS_ACCEPT"] = true
-MNIST.traindata()
+MNIST(Float32, :train)
 
 function reshape_data(X::AbstractArray{<:Real, 3})
     s = size(X)
@@ -28,8 +28,8 @@ function train_or_load!(file_name, m, args...; force=false, kwargs...)
 end
 
 function load_data(dataset; T=Float32, onehot=false, classes=0:9)
-    X_train, y_train = dataset.traindata(T)
-    X_test, y_test = dataset.testdata(T)
+    X_train, y_train = dataset(T, :train)[:]
+    X_test, y_test = dataset(T, :test)[:]
 
     X_train = reshape_data(X_train)
     X_test = reshape_data(X_test)

docs/src/lecture_11/iris.md

@@ -62,6 +62,8 @@ Because there are ``3`` classes and ``120`` samples in the training set, it retu
 We access the neural network parameters by using `params(m)`. We can select the second layer of `m` by `m[2]`. Since the second layer has ``5 `` input and ``3`` output neurons, its parameters are a matrix of size ``3\times 5`` and a vector of length ``3``. The parameters `params(m[2])` are a tuple of the matrix and the vector. This also implies that the parameters are initialized randomly, and we do not need to take care of it. We can easily modify any parameters.
 
 ```@example iris
+using Flux: params
+
 params(m[2])[2] .= [-1;0;1]
 
 nothing # hide

docs/src/lecture_11/nn.md

@@ -5,7 +5,7 @@ using MLDatasets
 
 Core.eval(Main, :(using Flux)) # hide
 ENV["DATADEPS_ALWAYS_ACCEPT"] = true
-X_train = MNIST.traindata()[1]
+X_train = MNIST(Float32, :train)[:][1]
 
 # imageplot(1 .- X_train, 1:15; nrows = 3, size=(800,480))
 
@@ -37,8 +37,8 @@ We use the package [MLDatasets](https://juliaml.github.io/MLDatasets.jl/stable/)
 using MLDatasets
 
 T = Float32
-X_train, y_train = MLDatasets.MNIST.traindata(T)
-X_test, y_test = MLDatasets.MNIST.testdata(T)
+X_train, y_train = MLDatasets.MNIST(T, :train)[:]
+X_test, y_test = MLDatasets.MNIST(T, :test)[:]
 
 nothing # hide
 ```
@@ -89,7 +89,6 @@ using Plots
 using ImageInspector
 
 imageplot(1 .- X_train, inds; nrows=3, size=(800,480))
-
 savefig("mnist_intro2.svg") # hide
 ```
 
@@ -161,8 +160,8 @@ using Flux
 using Flux: onehotbatch, onecold
 
 function load_data(dataset; T=Float32, onehot=false, classes=0:9)
-    X_train, y_train = dataset.traindata(T)
-    X_test, y_test = dataset.testdata(T)
+    X_train, y_train = dataset(T, :train)[:]
+    X_test, y_test = dataset(T, :test)[:]
 
     X_train = reshape_data(X_train)
     X_test = reshape_data(X_test)
@@ -363,6 +362,7 @@ To build the objective ``L``, we first specify the prediction function ``\operat
 
 ```@example nn
 using Random
+using Flux: flatten
 
 Random.seed!(666)
 m = Chain(
@@ -392,6 +392,7 @@ We now write the function `train_model!` to train the neural network `m`. Since
 
 ```@example nn
 using BSON
+using Flux: params
 
 function train_model!(m, L, X, y;
         opt = Descent(0.1),