fix datasetdict

Author: CarloLucibello

CondaPkg.toml

@@ -3,8 +3,8 @@ channels = ["conda-forge"]
 [deps]
 # h5py = ""
 # pillow = ">=9.1, <10"
-# numpy = ">=1.20, <2"
-pillow = ""
 datasets = ">=2.12, <3"
+numpy = ">=1.20, <2"
+pillow = ""
 
 # pyarrow = "==6.0.0"

src/dataset.jl

@@ -107,7 +107,7 @@ version of [`with_format`](@ref).
 """
 function set_format!(ds::Dataset, format)
     if format == "julia"
-        # ds.pyds.set_format("numpy")
+        ds.pyds.reset_format() # or d.pyd.set_format("python")
         ds.jltransform = py2jl
     else
         ds.pyds.set_format(format)

src/datasetdict.jl

@@ -102,7 +102,7 @@ version of [`with_format`](@ref).
 """
 function set_format!(d::DatasetDict, format)
     if format == "julia"
-        d.pyd.set_format("numpy")
+        d.pyd.reset_format()
         d.jltransform = py2jl
     else
         d.pyd.set_format(format)

src/transforms.jl

@@ -21,6 +21,7 @@ function _pyconvert(x::Py)
     end
 end
 
+# Do nothing on a non-Py object.
 _pyconvert(x) = x
 
 """
@@ -30,6 +31,7 @@ Convert Python types to Julia types applying `pyconvert` recursively.
 """
 py2jl
 
+# py2jl recurses through pycanonicalize and converts through _pyconvert
 py2jl(x) = pycanonicalize(_pyconvert(x))
 
 pycanonicalize(x) = x

test.jl

@@ -0,0 +1,4 @@
+using Test
+tmnist = load_dataset("mnist", split="test").with_format("julia")
+@test size(tmnist[1]["image"]) == (28, 28)
+

test/dataset.jl

@@ -74,7 +74,6 @@ end
     @test x isa Dict
     @test x["label"] == -1
     @test x["idx"] == 0
-    @show x["premise"] |> typeof
     @test x["premise"] isa AbstractString
     @test x["premise"] == "The cat sat on the mat."
     @test x["hypothesis"] isa AbstractString

test/datasetdict.jl

@@ -19,12 +19,12 @@ end
 @testset "with_format(julia)" begin
     d = with_format(mnist, "julia")
     ds = d["test"]
-    @test ds.format["type"] == "numpy"
+    @test ds.format["type"] == nothing
     x = ds[1]
     @test x isa Dict
     @test x["label"] isa Int
     @test x["label"] == 7
-    @test x["image"] isa Matrix{UInt8}
+    @test x["image"] isa AbstractMatrix{<:Gray}
     @test size(x["image"]) == (28, 28)
 end