Support onnx export with previous OpSet version (#5176)

* initial checkin

* fix

* fix build error

* change exception capture method

* print

* seperate DEBUG mode in test

* fix one build error

* use throw instead of assert for consistent test

* apply exception to all

* more fix

* review comments

* review comments2

* review part of comments

* review comments

* resolve comments

* revert files

* add description

src/Microsoft.ML.Data/Model/Onnx/OnnxContext.cs

@@ -43,6 +43,12 @@ internal abstract class OnnxContext
         /// <returns>Whether the column is mapped in this context</returns>
         public abstract bool ContainsColumn(string colName);
 
+        /// <summary>
+        /// Check the required OpSet version satisfies our requirement
+        /// </summary>
+        /// <returns></returns>
+        public abstract void CheckOpSetVersion(int thisTransformerMinumumOpSetVersion, string registerTransformerName);
+
         /// <summary>
         /// Stops tracking a column.
         /// </summary>

src/Microsoft.ML.Data/Prediction/Calibrator.cs

@@ -1745,6 +1745,9 @@ bool ISingleCanSaveOnnx.SaveAsOnnx(OnnxContext ctx, string[] scoreProbablityColu
             _host.CheckValue(scoreProbablityColumnNames, nameof(scoreProbablityColumnNames));
             _host.Check(Utils.Size(scoreProbablityColumnNames) == 2);
 
+            const int minimumOpSetVersion = 9;
+            ctx.CheckOpSetVersion(minimumOpSetVersion, "PlattCalibrator");
+
             // The Affine operator is no longer supported in the v11 opset.
             // So we have to decompose it using Mul and Add
             string opType = "Mul";

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

@@ -902,6 +902,9 @@ public void SaveAsOnnx(OnnxContext ctx)
                 Host.CheckValue(ctx, nameof(ctx));
                 Contracts.Assert(CanSaveOnnx(ctx));
 
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 for (int iinfo = 0; iinfo < _columns.Length; ++iinfo)
                 {
                     var colInfo = _parent._columns[iinfo];

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

@@ -237,6 +237,9 @@ protected override DataViewSchema.DetachedColumn[] GetOutputColumnsCore()
 
             public void SaveAsOnnx(OnnxContext ctx)
             {
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 var opType = "Identity";
 
                 foreach (var column in _columns)

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

@@ -732,6 +732,9 @@ IDataTransform ITransformTemplate.ApplyToData(IHostEnvironment env, IDataView ne
 
             public void SaveAsOnnx(OnnxContext ctx)
             {
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 var outputToInputMap = _mapper.OutputToInputMap;
                 for(int i = 0; i < outputToInputMap.Length; i++)
                 {

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

@@ -1353,6 +1353,9 @@ private void AddMetaKeyValues(int i, DataViewSchema.Annotations.Builder builder)
 
             private bool SaveAsOnnxCore(OnnxContext ctx, int iinfo, string srcVariable, string dstVariable)
             {
+                const int minimumOpSetVersion = 11;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 string castOutput;
                 string isGreaterThanZeroOutput = "";
                 OnnxNode castNode;

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

@@ -500,6 +500,9 @@ public override JToken SavePfa(BoundPfaContext ctx, JToken srcToken)
 
                 public override bool SaveOnnx(OnnxContext ctx, string srcVariableName, string dstVariableName)
                 {
+                    const int minimumOpSetVersion = 9;
+                    ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                     string opType;
 
                     // Onnx expects the input keys to be int64s. But the input data can come from an ML.NET node that

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

@@ -689,6 +689,9 @@ private JToken SaveAsPfaCore(BoundPfaContext ctx, int iinfo, ColInfo info, JToke
 
             private void SaveAsOnnxCore(OnnxContext ctx, int iinfo, ColInfo info, string srcVariableName, string dstVariableName)
             {
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 var dim = info.TypeSrc.GetValueCount();
 
                 string opType = "Cast";

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

@@ -822,6 +822,9 @@ private bool SaveAsOnnxCore(OnnxContext ctx, int iinfo, ColumnOptions info, stri
                 Contracts.Assert(_parent.Columns[iinfo] == info);
                 Contracts.Assert(CanSaveOnnx(ctx));
 
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 int valueCount = info.InputType.GetValueCount();
                 if (valueCount == 0)
                     return false;

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

@@ -895,6 +895,9 @@ public void SaveAsOnnx(OnnxContext ctx)
 
             public bool SaveAsOnnxCore(OnnxContext ctx, int iinfo, string srcVariableName, string dstVariableName)
             {
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 string opType;
                 var slots = _slotDropper[iinfo].GetPreservedSlots();
                 // vector column is not suppressed

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

@@ -497,6 +497,9 @@ public void SaveAsOnnx(OnnxContext ctx)
 
             private bool SaveAsOnnxCore(OnnxContext ctx, int iinfo, string srcVariableName, string dstVariableName)
             {
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 var opType = "Cast";
                 var node = ctx.CreateNode(opType, srcVariableName, dstVariableName, ctx.GetNodeName(opType), "");
                 var t = _parent._columns[iinfo].OutputKind.ToInternalDataKind().ToType();

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

@@ -812,6 +812,9 @@ private void CastInputToFloat<T>(OnnxContext ctx, out OnnxNode node, out long[]
 
             private bool SaveAsOnnxCore(OnnxContext ctx, int iinfo, ColInfo info, string srcVariableName, string dstVariableName)
             {
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 OnnxNode node;
                 long[] termIds;
                 string opType = "LabelEncoder";

src/Microsoft.ML.FastTree/FastTree.cs

@@ -3032,6 +3032,9 @@ private protected virtual bool SaveAsOnnx(OnnxContext ctx, string[] outputNames,
             Host.CheckValue(ctx, nameof(ctx));
             Host.Check(Utils.Size(outputNames) >= 1);
 
+            const int minimumOpSetVersion = 9;
+            ctx.CheckOpSetVersion(minimumOpSetVersion, "TreeEnsembleModelParameters");
+
             //Nodes.
             var nodesTreeids = new List<long>();
             var nodesIds = new List<long>();

src/Microsoft.ML.FastTree/FastTreeTweedie.cs

@@ -534,6 +534,9 @@ internal static FastTreeTweedieModelParameters Create(IHostEnvironment env, Mode
 
         bool ISingleCanSaveOnnx.SaveAsOnnx(OnnxContext ctx, string[] outputNames, string featureColumn)
         {
+            const int minimumOpSetVersion = 9;
+            ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
             // Mapping score to prediction
             var fastTreeOutput = ctx.AddIntermediateVariable(null, "FastTreeOutput", true);
             base.SaveAsOnnx(ctx, new[] { fastTreeOutput }, featureColumn);

src/Microsoft.ML.FastTree/RandomForestRegression.cs

@@ -213,6 +213,9 @@ internal static FastForestRegressionModelParameters Create(IHostEnvironment env,
 
         bool ISingleCanSaveOnnx.SaveAsOnnx(OnnxContext ctx, string[] outputNames, string featureColumn)
         {
+            const int minimumOpSetVersion = 9;
+            ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
             // Mapping score to prediction
             var fastTreeOutput = ctx.AddIntermediateVariable(null, "FastTreeOutput", true);
             var numTrees = ctx.AddInitializer((float)TrainedEnsemble.NumTrees, "NumTrees");

src/Microsoft.ML.KMeansClustering/KMeansModelParameters.cs

@@ -316,6 +316,9 @@ bool ISingleCanSaveOnnx.SaveAsOnnx(OnnxContext ctx, string[] outputNames, string
             //                                                                   v
             //                                           L [l] <--- ArgMin <---  Y [l, k]
 
+            const int minimumOpSetVersion = 9;
+            ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
             // Allocate C, which is a constant tensor in prediction phase
             var shapeC = new long[] { _centroids.Length, _centroids[0].Length };
             var tensorC = new List<float>();

src/Microsoft.ML.Mkl.Components/VectorWhitening.cs

@@ -641,6 +641,9 @@ public void SaveAsOnnx(OnnxContext ctx)
 
             private void SaveAsOnnxCore(OnnxContext ctx, int iinfo, string srcVariableName, string dstVariableName)
             {
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 var model = _parent._models[iinfo];
                 int dimension = _srcTypes[iinfo].GetValueCount();
                 Host.Assert(model.Length == dimension * dimension);

src/Microsoft.ML.OnnxConverter/OnnxContextImpl.cs

@@ -14,6 +14,8 @@ namespace Microsoft.ML.Model.OnnxConverter
     /// </summary>
     internal sealed class OnnxContextImpl : OnnxContext
     {
+        private const int CurrentOpSetVersion = 12;
+        private const int MinimumOpSetVersion = 9;
         private readonly List<OnnxCSharpToProtoWrapper.NodeProto> _nodes;
         private readonly List<OnnxUtils.ModelArgs> _inputs;
         // The map from IDataView column names to variable names.
@@ -32,9 +34,10 @@ internal sealed class OnnxContextImpl : OnnxContext
         private readonly string _producerVersion;
         private readonly long _modelVersion;
         private readonly OnnxVersion _onnxVersion;
+        private readonly int _opSetVersion;
 
         public OnnxContextImpl(IHostEnvironment env, string name, string producerName,
-            string producerVersion, long modelVersion, string domain, OnnxVersion onnxVersion)
+            string producerVersion, long modelVersion, string domain, OnnxVersion onnxVersion, int opSetVersion = CurrentOpSetVersion)
         {
             Contracts.CheckValue(env, nameof(env));
             _host = env.Register(nameof(OnnxContext));
@@ -55,6 +58,9 @@ public OnnxContextImpl(IHostEnvironment env, string name, string producerName,
             _modelVersion = modelVersion;
             _domain = domain;
             _onnxVersion = onnxVersion;
+            _opSetVersion = opSetVersion <= CurrentOpSetVersion ?
+                            opSetVersion >= MinimumOpSetVersion ? opSetVersion : throw _host.ExceptParam(nameof(opSetVersion), $"Requested OpSet version {opSetVersion} is lower than the minimum required OpSet version {MinimumOpSetVersion}") :
+                            throw _host.ExceptParam(nameof(opSetVersion), $"Requested OpSet version {opSetVersion} is higher than the current most updated OpSet version {CurrentOpSetVersion}");
         }
 
         public override bool ContainsColumn(string colName) => _columnNameMap.ContainsKey(colName);
@@ -127,6 +133,12 @@ public override string GetNodeName(string prefix)
             return GetUniqueName(prefix, _nodeNames.Contains);
         }
 
+        public override void CheckOpSetVersion(int thisTransformerMinumumOpSetVersion, string registerTransformerName)
+        {
+            if (_opSetVersion < thisTransformerMinumumOpSetVersion)
+                throw _host.ExceptParam(nameof(thisTransformerMinumumOpSetVersion), $"Requested OpSet version {_opSetVersion} is lower than {registerTransformerName}'s minimum OpSet version requirement: {thisTransformerMinumumOpSetVersion}");
+        }
+
         /// <summary>
         /// Adds a node to the node list of the graph.
         /// </summary>
@@ -409,7 +421,7 @@ public override string AddInitializer(IEnumerable<ulong> values, bool isUint64,
         /// Makes the ONNX model based on the context.
         /// </summary>
         public OnnxCSharpToProtoWrapper.ModelProto MakeModel()
-            => OnnxUtils.MakeModel(_nodes, _producerName, _name, _domain, _producerVersion, _modelVersion, _inputs, _outputs, _intermediateValues, _initializers);
+            => OnnxUtils.MakeModel(_nodes, _producerName, _name, _domain, _producerVersion, _modelVersion, _opSetVersion, _inputs, _outputs, _intermediateValues, _initializers);
 
         /// <summary>
         /// Return either "Experimental" or "Stable". The string "Experimental" indicates that some experimental features which are

src/Microsoft.ML.OnnxConverter/OnnxExportExtensions.cs

@@ -7,12 +7,23 @@
 using Google.Protobuf;
 using Microsoft.ML.Data;
 using Microsoft.ML.Model.OnnxConverter;
+using Microsoft.ML.Runtime;
 using static Microsoft.ML.Model.OnnxConverter.OnnxCSharpToProtoWrapper;
 
 namespace Microsoft.ML
 {
     public static class OnnxExportExtensions
     {
+        private static ModelProto ConvertToOnnxProtobufCore(IHostEnvironment env, OnnxContextImpl ctx, ITransformer transform, IDataView inputData)
+        {
+            var outputData = transform.Transform(inputData);
+            LinkedList<ITransformCanSaveOnnx> transforms = null;
+            using (var ch = env.Start("ONNX conversion"))
+            {
+                SaveOnnxCommand.GetPipe(ctx, ch, outputData, out IDataView root, out IDataView sink, out transforms);
+                return SaveOnnxCommand.ConvertTransformListToOnnxModel(ctx, ch, root, sink, transforms, null, null);
+            }
+        }
 
         /// <summary>
         /// Convert the specified <see cref="ITransformer"/> to ONNX format. Note that ONNX uses Google's Protobuf so the returned value is a Protobuf object.
@@ -26,13 +37,23 @@ internal static ModelProto ConvertToOnnxProtobuf(this ModelOperationsCatalog cat
         {
             var env = catalog.GetEnvironment();
             var ctx = new OnnxContextImpl(env, "model", "ML.NET", "0", 0, "machinelearning.dotnet", OnnxVersion.Stable);
-            var outputData = transform.Transform(inputData);
-            LinkedList<ITransformCanSaveOnnx> transforms = null;
-            using (var ch = env.Start("ONNX conversion"))
-            {
-                SaveOnnxCommand.GetPipe(ctx, ch, outputData, out IDataView root, out IDataView sink, out transforms);
-                return SaveOnnxCommand.ConvertTransformListToOnnxModel(ctx, ch, root, sink, transforms, null, null);
-            }
+            return ConvertToOnnxProtobufCore(env, ctx, transform, inputData);
+        }
+
+        /// <summary>
+        /// Convert the specified <see cref="ITransformer"/> to ONNX format. Note that ONNX uses Google's Protobuf so the returned value is a Protobuf object.
+        /// </summary>
+        /// <param name="catalog">The class that <see cref="ConvertToOnnxProtobuf(ModelOperationsCatalog, ITransformer, IDataView, int)"/> attached to.</param>
+        /// <param name="transform">The <see cref="ITransformer"/> that will be converted into ONNX format.</param>
+        /// <param name="inputData">The input of the specified transform.</param>
+        /// <param name="opSetVersion">The OpSet version to use for exporting the model. This value must be greater than or equal to 9 and less than or equal to 12</param>
+        /// <returns>An ONNX model equivalent to the converted ML.NET model.</returns>
+        [BestFriend]
+        internal static ModelProto ConvertToOnnxProtobuf(this ModelOperationsCatalog catalog, ITransformer transform, IDataView inputData, int opSetVersion)
+        {
+            var env = catalog.GetEnvironment();
+            var ctx = new OnnxContextImpl(env, "model", "ML.NET", "0", 0, "machinelearning.dotnet", OnnxVersion.Stable, opSetVersion);
+            return ConvertToOnnxProtobufCore(env, ctx, transform, inputData);
         }
 
         /// <summary>
@@ -45,5 +66,17 @@ internal static ModelProto ConvertToOnnxProtobuf(this ModelOperationsCatalog cat
         /// <returns>An ONNX model equivalent to the converted ML.NET model.</returns>
         public static void ConvertToOnnx(this ModelOperationsCatalog catalog, ITransformer transform, IDataView inputData, Stream stream) =>
             ConvertToOnnxProtobuf(catalog, transform, inputData).WriteTo(stream);
+
+        /// <summary>
+        /// Convert the specified <see cref="ITransformer"/> to ONNX format and writes to a stream.
+        /// </summary>
+        /// <param name="catalog">The class that <see cref="ConvertToOnnx(ModelOperationsCatalog, ITransformer, IDataView, int, Stream)"/> attached to.</param>
+        /// <param name="transform">The <see cref="ITransformer"/> that will be converted into ONNX format.</param>
+        /// <param name="inputData">The input of the specified transform.</param>
+        /// <param name="opSetVersion">The OpSet version to use for exporting the model. This value must be greater than or equal to 9 and less than or equal to 12</param>
+        /// <param name="stream">The stream to write the protobuf model to.</param>
+        /// <returns>An ONNX model equivalent to the converted ML.NET model.</returns>
+        public static void ConvertToOnnx(this ModelOperationsCatalog catalog, ITransformer transform, IDataView inputData, int opSetVersion, Stream stream) =>
+            ConvertToOnnxProtobuf(catalog, transform, inputData, opSetVersion).WriteTo(stream);
     }
 }

src/Microsoft.ML.OnnxConverter/OnnxUtils.cs

@@ -285,7 +285,7 @@ public ModelArgs(string name, TensorProto.Types.DataType dataType, List<long> di
         }
 
         public static ModelProto MakeModel(List<NodeProto> nodes, string producerName, string name,
-            string domain, string producerVersion, long modelVersion, List<ModelArgs> inputs,
+            string domain, string producerVersion, long modelVersion, int opSetVersion, List<ModelArgs> inputs,
             List<ModelArgs> outputs, List<ModelArgs> intermediateValues, List<TensorProto> initializers)
         {
             Contracts.CheckValue(nodes, nameof(nodes));
@@ -305,7 +305,7 @@ public static ModelProto MakeModel(List<NodeProto> nodes, string producerName, s
             model.IrVersion = (long)OnnxCSharpToProtoWrapper.Version.IrVersion;
             model.ModelVersion = modelVersion;
             model.OpsetImport.Add(new OperatorSetIdProto() { Domain = "ai.onnx.ml", Version = 2 });
-            model.OpsetImport.Add(new OperatorSetIdProto() { Domain = "", Version = 11 });
+            model.OpsetImport.Add(new OperatorSetIdProto() { Domain = "", Version = opSetVersion });
             model.Graph = new GraphProto();
             var graph = model.Graph;
             graph.Node.Add(nodes);

src/Microsoft.ML.PCA/PcaTransformer.cs

@@ -625,6 +625,9 @@ private void SaveAsOnnxCore(OnnxContext ctx, int iinfo, string srcVariableName,
             {
                 Host.CheckValue(ctx, nameof(ctx));
 
+                const int minimumOpSetVersion = 9;
+                ctx.CheckOpSetVersion(minimumOpSetVersion, LoaderSignature);
+
                 TransformInfo transformInfo = _parent._transformInfos[iinfo];
 
                 // When the transformer is loaded from a model file,

src/Microsoft.ML.StandardTrainers/Standard/LinearModelParameters.cs

@@ -137,6 +137,10 @@ private protected virtual bool SaveAsOnnx(OnnxContext ctx, string[] outputs, str
         {
             Host.CheckValue(ctx, nameof(ctx));
             Host.Check(Utils.Size(outputs) >= 1);
+
+            const int minimumOpSetVersion = 9;
+            ctx.CheckOpSetVersion(minimumOpSetVersion, "LinearModel");
+
             string opType = "LinearRegressor";
             string scoreVarName = (Utils.Size(outputs) >= 2) ? outputs[1] : outputs[0]; // Get Score from PredictedLabel and/or Score columns
             var node = ctx.CreateNode(opType, new[] { featureColumn }, new[] { scoreVarName }, ctx.GetNodeName(opType));

src/Microsoft.ML.StandardTrainers/Standard/LogisticRegression/MulticlassLogisticRegression.cs

@@ -978,6 +978,10 @@ private JToken SaveAsPfaCore(BoundPfaContext ctx, JToken input)
         private bool SaveAsOnnxCore(OnnxContext ctx, string[] outputs, string featureColumn)
         {
             Host.CheckValue(ctx, nameof(ctx));
+
+            const int minimumOpSetVersion = 9;
+            ctx.CheckOpSetVersion(minimumOpSetVersion, "MultiClassLogisticRegression");
+
             Host.Assert(outputs[0] == DefaultColumnNames.PredictedLabel);
             Host.Assert(outputs[1] == DefaultColumnNames.Score);
 

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

@@ -421,6 +421,9 @@ ValueMapper<TIn, TOut> IValueMapper.GetMapper<TIn, TOut>()
         /// </summary>
         bool ISingleCanSaveOnnx.SaveAsOnnx(OnnxContext ctx, string[] outputNames, string featureColumn)
         {
+            const int minimumOpSetVersion = 9;
+            ctx.CheckOpSetVersion(minimumOpSetVersion, "MulticlassNaiveBayes");
+
             float[] featureHistogram = new float[_featureHistogram[0].Length * _labelHistogram.Length];
             float[] labelHistogramExpanded = new float[_featureHistogram[0].Length * _labelHistogram.Length];