adding baselines

Author: Keren Fuentes

src/Microsoft.ML.Data/Transforms/ColumnConcatenatingTransformer.cs

@@ -896,15 +896,14 @@ public void SaveAsOnnx(OnnxContext ctx)
                 Host.CheckValue(ctx, nameof(ctx));
                 Contracts.Assert(CanSaveOnnx(ctx));
 
-                string opType = "FeatureVectorizer";
                 for (int iinfo = 0; iinfo < _columns.Length; ++iinfo)
                 {
                     var colInfo = _parent._columns[iinfo];
                     var boundCol = _columns[iinfo];
 
                     string outName = colInfo.Name;
                     var outColType = boundCol.OutputType;
-                    if (!outColType.IsKnownSize)
+                    if ((!outColType.IsKnownSize) || (!(outColType.GetItemType() is NumberDataViewType)))
                     {
                         ctx.RemoveColumn(outName, false);
                         continue;
@@ -925,15 +924,18 @@ public void SaveAsOnnx(OnnxContext ctx)
                             InputSchema[srcIndex].Type.GetValueCount()));
                     }
 
-                    var dstVariableName = ctx.AddIntermediateVariable(outColType, outName, true);
-                    var vectorizerOutputName = ctx.AddIntermediateVariable(null, "VectorFeaturizerOutput", true);
+                    string opType = "FeatureVectorizer";
+                    int outVectorSize = (int)inputList.Sum(x => x.Value);
+                    var vectorizerOutputType = new VectorDataViewType(NumberDataViewType.Single, outVectorSize);
+                    var vectorizerOutputName = ctx.AddIntermediateVariable(vectorizerOutputType, "VectorFeaturizerOutput");
                     var node = ctx.CreateNode(opType, inputList.Select(t => t.Key),
                         new[] { vectorizerOutputName }, ctx.GetNodeName(opType));
                     node.AddAttribute("inputdimensions", inputList.Select(x => x.Value));
 
                     opType = "Cast";
+                    var dstVectorType = new VectorDataViewType(outColType.GetItemType() as PrimitiveDataViewType, outVectorSize);
+                    var dstVariableName = ctx.AddIntermediateVariable(dstVectorType, outName);
                     var castNode = ctx.CreateNode(opType, vectorizerOutputName, dstVariableName, ctx.GetNodeName(opType), "");
-                    var t = InternalDataKindExtensions.ToInternalDataKind(DataKind.Double).ToType();
                     castNode.AddAttribute("to", outColType.ItemType.RawType);
                 }
             }

test/BaselineOutput/Common/Onnx/BinaryClassification/BreastCancer/ExcludeVariablesInOnnxConversion.txt

@@ -176,14 +176,35 @@
           "F1",
           "F22"
         ],
+        "output": [
+          "VectorFeaturizerOutput"
+        ],
+        "name": "FeatureVectorizer",
+        "opType": "FeatureVectorizer",
+        "attribute": [
+          {
+            "name": "inputdimensions",
+            "ints": [
+              "1",
+              "10"
+            ],
+            "type": "INTS"
+          }
+        ],
+        "domain": "ai.onnx.ml"
+      },
+      {
+        "input": [
+          "VectorFeaturizerOutput"
+        ],
         "output": [
           "Features"
         ],
-        "name": "Concat",
-        "opType": "Concat",
+        "name": "Cast1",
+        "opType": "Cast",
         "attribute": [
           {
-            "name": "axis",
+            "name": "to",
             "i": "1",
             "type": "INT"
           }
@@ -431,7 +452,7 @@
         "output": [
           "PredictedLabel"
         ],
-        "name": "Cast1",
+        "name": "Cast2",
         "opType": "Cast",
         "attribute": [
           {
@@ -638,6 +659,24 @@
           }
         }
       },
+      {
+        "name": "VectorFeaturizerOutput",
+        "type": {
+          "tensorType": {
+            "elemType": 1,
+            "shape": {
+              "dim": [
+                {
+                  "dimValue": "-1"
+                },
+                {
+                  "dimValue": "11"
+                }
+              ]
+            }
+          }
+        }
+      },
       {
         "name": "Features",
         "type": {

test/BaselineOutput/Common/Onnx/BinaryClassification/BreastCancer/ModelWithLessIO.txt

@@ -125,14 +125,35 @@
           "F1",
           "F21"
         ],
+        "output": [
+          "VectorFeaturizerOutput"
+        ],
+        "name": "FeatureVectorizer",
+        "opType": "FeatureVectorizer",
+        "attribute": [
+          {
+            "name": "inputdimensions",
+            "ints": [
+              "8",
+              "9"
+            ],
+            "type": "INTS"
+          }
+        ],
+        "domain": "ai.onnx.ml"
+      },
+      {
+        "input": [
+          "VectorFeaturizerOutput"
+        ],
         "output": [
           "Features"
         ],
-        "name": "Concat",
-        "opType": "Concat",
+        "name": "Cast1",
+        "opType": "Cast",
         "attribute": [
           {
-            "name": "axis",
+            "name": "to",
             "i": "1",
             "type": "INT"
           }
@@ -757,7 +778,7 @@
         "output": [
           "PredictedLabel"
         ],
-        "name": "Cast1",
+        "name": "Cast2",
         "opType": "Cast",
         "attribute": [
           {
@@ -946,6 +967,24 @@
           }
         }
       },
+      {
+        "name": "VectorFeaturizerOutput",
+        "type": {
+          "tensorType": {
+            "elemType": 1,
+            "shape": {
+              "dim": [
+                {
+                  "dimValue": "-1"
+                },
+                {
+                  "dimValue": "17"
+                }
+              ]
+            }
+          }
+        }
+      },
       {
         "name": "Features",
         "type": {

test/BaselineOutput/Common/Onnx/BinaryClassification/BreastCancer/OneHotBagPipeline.txt

@@ -176,14 +176,35 @@
           "F1",
           "F22"
         ],
+        "output": [
+          "VectorFeaturizerOutput"
+        ],
+        "name": "FeatureVectorizer",
+        "opType": "FeatureVectorizer",
+        "attribute": [
+          {
+            "name": "inputdimensions",
+            "ints": [
+              "1",
+              "10"
+            ],
+            "type": "INTS"
+          }
+        ],
+        "domain": "ai.onnx.ml"
+      },
+      {
+        "input": [
+          "VectorFeaturizerOutput"
+        ],
         "output": [
           "Features"
         ],
-        "name": "Concat",
-        "opType": "Concat",
+        "name": "Cast1",
+        "opType": "Cast",
         "attribute": [
           {
-            "name": "axis",
+            "name": "to",
             "i": "1",
             "type": "INT"
           }
@@ -384,7 +405,7 @@
         "output": [
           "PredictedLabel"
         ],
-        "name": "Cast1",
+        "name": "Cast2",
         "opType": "Cast",
         "attribute": [
           {
@@ -871,6 +892,24 @@
           }
         }
       },
+      {
+        "name": "VectorFeaturizerOutput",
+        "type": {
+          "tensorType": {
+            "elemType": 1,
+            "shape": {
+              "dim": [
+                {
+                  "dimValue": "-1"
+                },
+                {
+                  "dimValue": "11"
+                }
+              ]
+            }
+          }
+        }
+      },
       {
         "name": "Features",
         "type": {