TokenizingByCharacters export to Onnx

src/Microsoft.ML.StandardTrainers/Standard/MulticlassClassification/MulticlassNaiveBayesTrainer.cs

@@ -43,7 +43,7 @@ namespace Microsoft.ML.Trainers
     /// | Is normalization required? | Yes |
     /// | Is caching required? | No |
     /// | Required NuGet in addition to Microsoft.ML | None |
-    /// | Exportable to ONNX | No |
+    /// | Exportable to ONNX | Yes |
     ///
     /// ### Training Algorithm Details
     /// [Naive Bayes](https://en.wikipedia.org/wiki/Naive_Bayes_classifier)

src/Microsoft.ML.Transforms/Text/TextNormalizing.cs

@@ -442,7 +442,7 @@ private bool IsCombiningDiacritic(char ch)
     /// | Does this estimator need to look at the data to train its parameters? | No |
     /// | Input column data type | Scalar or Vector of [Text](xref:Microsoft.ML.Data.TextDataViewType)|
     /// | Output column data type | Scalar or variable-sized Vector of [Text](xref:Microsoft.ML.Data.TextDataViewType)|
-    /// | Exportable to ONNX | No |
+    /// | Exportable to ONNX | Yes |
     ///
     /// The resulting <xref:Microsoft.ML.Transforms.Text.TextNormalizingTransformer> creates a new column, named as specified
     /// in the output column name parameters, and normalizes the textual input data by changing case, removing diacritical marks,

src/Microsoft.ML.Transforms/Text/TokenizingByCharacters.cs

@@ -11,6 +11,7 @@
 using Microsoft.ML.CommandLine;
 using Microsoft.ML.Data;
 using Microsoft.ML.Internal.Utilities;
+using Microsoft.ML.Model.OnnxConverter;
 using Microsoft.ML.Runtime;
 using Microsoft.ML.Transforms.Text;
 
@@ -184,7 +185,7 @@ private static IRowMapper Create(IHostEnvironment env, ModelLoadContext ctx, Dat
 
         private protected override IRowMapper MakeRowMapper(DataViewSchema schema) => new Mapper(this, schema);
 
-        private sealed class Mapper : OneToOneMapperBase
+        private sealed class Mapper : OneToOneMapperBase, ISaveAsOnnx
         {
             private readonly DataViewType _type;
             private readonly TokenizingByCharactersTransformer _parent;
@@ -204,6 +205,54 @@ public Mapper(TokenizingByCharactersTransformer parent, DataViewSchema inputSche
                     _isSourceVector[i] = inputSchema[_parent.ColumnPairs[i].inputColumnName].Type is VectorDataViewType;
             }
 
+            public bool CanSaveOnnx(OnnxContext ctx) => true;
+
+            public void SaveAsOnnx(OnnxContext ctx)
+            {
+                Host.CheckValue(ctx, nameof(ctx));
+                for (int iinfo = 0; iinfo < _isSourceVector.Length; ++iinfo)
+                {
+                    string inputColumnName = _parent.ColumnPairs[iinfo].inputColumnName;
+                    if (!ctx.ContainsColumn(inputColumnName))
+                        continue;
+
+                    string outputColumnName = _parent.ColumnPairs[iinfo].outputColumnName;
+                    string srcVariableName = ctx.GetVariableName(inputColumnName);
+                    string dstVariableName = ctx.AddIntermediateVariable(_type, outputColumnName, true);
+                    SaveAsOnnxCore(ctx, srcVariableName, dstVariableName);
+                }
+            }
+
+            private void SaveAsOnnxCore(OnnxContext ctx, string srcVariableName, string dstVariableName)
+            {
+                string opType = "Tokenizer";
+                string tokenizerOutput = ctx.AddIntermediateVariable(null, "TokenizerOutput", true);
+                var node = ctx.CreateNode(opType, srcVariableName, tokenizerOutput, ctx.GetNodeName(opType), "com.microsoft");
+                node.AddAttribute("mark", _parent._useMarkerChars);
+                node.AddAttribute("mincharnum", 1);
+                node.AddAttribute("pad_value", "");
+                node.AddAttribute("separators", new string[] { "" });
+
+                opType = "Squeeze";
+                var squeezeOutput = ctx.AddIntermediateVariable(null, "SqueezeOutput", true);
+                node = ctx.CreateNode(opType, tokenizerOutput, squeezeOutput, ctx.GetNodeName(opType), "");
+                node.AddAttribute("axes", new long[] { 0 });
+
+                opType = "LabelEncoder";
+                var labelEncoderOutput = ctx.AddIntermediateVariable(null, "LabelEncoderOutput", true);
+                node = ctx.CreateNode(opType, squeezeOutput, labelEncoderOutput, ctx.GetNodeName(opType));
+
+                IEnumerable<string> charStrings = Enumerable.Range(0, 65535).Select(x => ((char)x).ToString());
+                IEnumerable<long> charValues = Enumerable.Range(0, 65535).Select(x => Convert.ToInt64(x)); ;
+                node.AddAttribute("keys_strings", charStrings);
+                node.AddAttribute("values_int64s", charValues);
+
+                opType = "Cast";
+                var castNode = ctx.CreateNode(opType, labelEncoderOutput, dstVariableName, ctx.GetNodeName(opType), "");
+                var t = InternalDataKindExtensions.ToInternalDataKind(DataKind.UInt16).ToType();
+                castNode.AddAttribute("to", t);
+            }
+
             protected override DataViewSchema.DetachedColumn[] GetOutputColumnsCore()
             {
                 var result = new DataViewSchema.DetachedColumn[_parent.ColumnPairs.Length];
@@ -558,7 +607,7 @@ private ValueGetter<VBuffer<ushort>> MakeGetterVec(DataViewRow input, int iinfo)
     /// | Does this estimator need to look at the data to train its parameters? | Yes |
     /// | Input column data type | Scalar or Vector of [Text](xref:Microsoft.ML.Data.TextDataViewType)  |
     /// | Output column data type | Variable-sized vector of [key](xref:Microsoft.ML.Data.KeyDataViewType) type. |
-    /// | Exportable to ONNX | No |
+    /// | Exportable to ONNX | Yes |
     ///
     /// The estimator tokenizes characters by splitting text into sequences of characters using a sliding window.
     /// During training, the estimator builds a key-value pair dictionary with the encountered sequences of characters.

test/Microsoft.ML.Tests/OnnxConversionTest.cs

@@ -921,6 +921,37 @@ public void WordEmbeddingsTest()
             Done();
         }
 
+        [Theory]
+        [CombinatorialData]
+        public void TokenizingByCharactersOnnxConversionTest(bool useMarkerCharacters)
+        {
+            var mlContext = new MLContext(seed: 1);
+            var dataPath = GetDataPath("wikipedia-detox-250-line-test.tsv");
+            var dataView = ML.Data.LoadFromTextFile(dataPath, new[] {
+                new TextLoader.Column("label", DataKind.Boolean, 0),
+                new TextLoader.Column("text", DataKind.String, 1)
+            }, hasHeader: true);
+            var pipeline = new TokenizingByCharactersEstimator(mlContext, useMarkerCharacters: useMarkerCharacters, columns: new[] { ("TokenizedText", "text") });
+            var model = pipeline.Fit(dataView);
+            var transformedData = model.Transform(dataView);
+            var onnxModel = mlContext.Model.ConvertToOnnxProtobuf(model, dataView);
+            // Compare model scores produced by ML.NET and ONNX's runtime. 
+            if (IsOnnxRuntimeSupported())
+            {
+                var onnxFileName = $"TokenizingByCharacters.onnx";
+                var onnxModelPath = GetOutputPath(onnxFileName);
+                SaveOnnxModel(onnxModel, onnxModelPath, null);
+                // Evaluate the saved ONNX model using the data used to train the ML.NET pipeline.
+                string[] inputNames = onnxModel.Graph.Input.Select(valueInfoProto => valueInfoProto.Name).ToArray();
+                string[] outputNames = onnxModel.Graph.Output.Select(valueInfoProto => valueInfoProto.Name).ToArray();
+                var onnxEstimator = mlContext.Transforms.ApplyOnnxModel(outputNames, inputNames, onnxModelPath);
+                var onnxTransformer = onnxEstimator.Fit(dataView);
+                var onnxResult = onnxTransformer.Transform(dataView);
+                CompareSelectedVectorColumns<UInt16>(transformedData.Schema[2].Name, outputNames[2], transformedData, onnxResult); //compare scores
+            }
+            Done();
+        }
+
         [Theory]
         // These are the supported conversions
         // ML.NET does not allow any conversions between signed and unsigned numeric types