Merge pull request #287 from FluxML/dev

Rebase entity-tutorial

Author: EssamWisam

Project.toml

@@ -1,7 +1,7 @@
 name = "MLJFlux"
 uuid = "094fc8d1-fd35-5302-93ea-dabda2abf845"
 authors = ["Anthony D. Blaom <anthony.blaom@gmail.com>", "Ayush Shridhar <ayush.shridhar1999@gmail.com>"]
-version = "0.6.0"
+version = "0.6.1"
 
 [deps]
 CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"

src/classifier.jl

@@ -40,7 +40,7 @@ function MLJModelInterface.predict(
 )
     chain, levels, ordinal_mappings, _ = fitresult
     Xnew = ordinal_encoder_transform(Xnew, ordinal_mappings)        # what if Xnew is a matrix
-    X = reformat(Xnew)
+    X = _f32(reformat(Xnew), 0)
     probs = vcat([chain(tomat(X[:, i]))' for i in 1:size(X, 2)]...)
     return MLJModelInterface.UnivariateFinite(levels, probs)
 end
@@ -69,7 +69,7 @@ function MLJModelInterface.predict(
 )
     chain, levels, ordinal_mappings, _ = fitresult
     Xnew = ordinal_encoder_transform(Xnew, ordinal_mappings)
-    X = reformat(Xnew)
+    X = _f32(reformat(Xnew), 0)
     probs = vec(chain(X))
     return MLJModelInterface.UnivariateFinite(levels, probs; augment = true)
 end

src/encoders.jl

@@ -67,7 +67,7 @@ function ordinal_encoder_transform(X, mapping_matrix)
             test_levels = levels(col)
             check_unkown_levels(train_levels, test_levels)
             level2scalar = mapping_matrix[ind]
-            new_col = [recode(unwrap.(col), level2scalar...)...]
+            new_col = unwrap.(recode(col, level2scalar...))
             push!(new_feats, new_col)
         else
             push!(new_feats, col)

src/regressor.jl

@@ -27,7 +27,7 @@ function MLJModelInterface.predict(model::NeuralNetworkRegressor,
     Xnew)
     chain, ordinal_mappings = fitresult[1], fitresult[3]
     Xnew = ordinal_encoder_transform(Xnew, ordinal_mappings)
-    Xnew_ = reformat(Xnew)
+    Xnew_ = _f32(reformat(Xnew), 0)
     return [chain(values.(tomat(Xnew_[:, i])))[1]
             for i in 1:size(Xnew_, 2)]
 end
@@ -74,7 +74,7 @@ function MLJModelInterface.predict(model::MultitargetNeuralNetworkRegressor,
     fitresult, Xnew)
     chain, target_column_names, ordinal_mappings, _ = fitresult
     Xnew = ordinal_encoder_transform(Xnew, ordinal_mappings)
-    X = reformat(Xnew)
+    X = _f32(reformat(Xnew), 0)
     ypred = [chain(values.(tomat(X[:, i])))
              for i in 1:size(X, 2)]
     output =

test/mlj_model_interface.jl

@@ -133,6 +133,13 @@ mutable struct LisasBuilder
   n1::Int
 end
 
+# UndefVarError accepts two inputs from julia > v"1.9"
+_UndefVarError(var, scope) = @static if VERSION < v"1.10"
+    UndefVarError(var)
+else
+    UndefVarError(var, scope)
+end
+
 @testset "builder errors and issue #237" begin
     # create a builder with an intentional flaw;
     # `Chains` is undefined - it should be `Chain`
@@ -153,7 +160,7 @@ end
     y = rand(Float32, 75)
     @test_logs(
         (:error, MLJFlux.ERR_BUILDER),
-        @test_throws UndefVarError(:Chains) MLJBase.fit(model, 0, X, y)
+        @test_throws _UndefVarError(:Chains, Flux) MLJBase.fit(model, 0, X, y)
     )
 end