Hide delegates, model parameters classes, move onFit to staticPIpe, get rid of trivial transformWrapper

Author: Ivan Matantsev

src/Microsoft.ML.Core/Data/InPredicate.cs

@@ -4,5 +4,6 @@
 
 namespace Microsoft.ML.Data
 {
-    public delegate bool InPredicate<T>(in T value);
+    [BestFriend]
+    internal delegate bool InPredicate<T>(in T value);
 }

src/Microsoft.ML.Data/DataLoadSave/TransformWrapper.cs

@@ -168,26 +168,4 @@ public SchemaShape GetOutputSchema(SchemaShape inputSchema)
             return SchemaShape.Create(transformer.GetOutputSchema(fakeSchema));
         }
     }
-
-    /// <summary>
-    /// Estimator for untrained wrapped transformers.
-    /// </summary>
-    public abstract class TrivialWrapperEstimator : TrivialEstimator<TransformWrapper>
-    {
-        protected TrivialWrapperEstimator(IHost host, TransformWrapper transformer)
-            : base(host, transformer)
-        {
-        }
-
-        /// <summary>
-        /// Returns the <see cref="SchemaShape"/> of the schema which will be produced by the transformer.
-        /// Used for schema propagation and verification in a pipeline.
-        /// </summary>
-        public override SchemaShape GetOutputSchema(SchemaShape inputSchema)
-        {
-            Host.CheckValue(inputSchema, nameof(inputSchema));
-            var fakeSchema = FakeSchemaFactory.Create(inputSchema);
-            return SchemaShape.Create(Transformer.GetOutputSchema(fakeSchema));
-        }
-    }
 }

src/Microsoft.ML.Data/Dirty/ILoss.cs

@@ -4,7 +4,6 @@
 
 using System;
 using Microsoft.ML.EntryPoints;
-using Float = System.Single;
 
 namespace Microsoft.ML
 {
@@ -17,12 +16,12 @@ public interface ILossFunction<in TOutput, in TLabel>
         Double Loss(TOutput output, TLabel label);
     }
 
-    public interface IScalarOutputLoss : ILossFunction<Float, Float>
+    public interface IScalarOutputLoss : ILossFunction<float, float>
     {
         /// <summary>
         /// Derivative of the loss function with respect to output
         /// </summary>
-        Float Derivative(Float output, Float label);
+        float Derivative(float output, float label);
     }
 
     [TlcModule.ComponentKind("RegressionLossFunction")]
@@ -46,10 +45,10 @@ public interface IClassificationLoss : IScalarOutputLoss
     /// <summary>
     /// Delegate signature for standardized classification loss functions.
     /// </summary>
-    public delegate void SignatureClassificationLoss();
+    internal delegate void SignatureClassificationLoss();
 
     /// <summary>
     /// Delegate signature for standardized regression loss functions.
     /// </summary>
-    public delegate void SignatureRegressionLoss();
+    internal delegate void SignatureRegressionLoss();
 }

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

@@ -159,25 +159,4 @@ public enum KeyValueOrder : byte
         /// </summary>
         Value = ValueToKeyMappingEstimator.SortOrder.Value
     }
-
-    /// <summary>
-    /// Information on the result of fitting a to-key transform.
-    /// </summary>
-    /// <typeparam name="T">The type of the values.</typeparam>
-    public sealed class ToKeyFitResult<T>
-    {
-        /// <summary>
-        /// For user defined delegates that accept instances of the containing type.
-        /// </summary>
-        /// <param name="result"></param>
-        public delegate void OnFit(ToKeyFitResult<T> result);
-
-        // At the moment this is empty. Once PR #863 clears, we can change this class to hold the output
-        // key-values metadata.
-
-        [BestFriend]
-        internal ToKeyFitResult(ValueToKeyMappingTransformer.TermMap map)
-        {
-        }
-    }
 }

src/Microsoft.ML.FastTree/GamClassification.cs

@@ -180,7 +180,7 @@ public sealed class BinaryClassificationGamModelParameters : GamModelParametersB
         /// <param name="featureToInputMap">A map from the feature shape functions (as described by the binUpperBounds and BinEffects)
         /// to the input feature. Used when the number of input features is different than the number of shape functions. Use default if all features have
         /// a shape function.</param>
-        public BinaryClassificationGamModelParameters(IHostEnvironment env,
+        internal BinaryClassificationGamModelParameters(IHostEnvironment env,
             double[][] binUpperBounds, double[][] binEffects, double intercept, int inputLength, int[] featureToInputMap)
             : base(env, LoaderSignature, binUpperBounds, binEffects, intercept, inputLength, featureToInputMap) { }
 

src/Microsoft.ML.FastTree/GamRegression.cs

@@ -126,7 +126,7 @@ public sealed class RegressionGamModelParameters : GamModelParametersBase
         /// <param name="featureToInputMap">A map from the feature shape functions (as described by the binUpperBounds and BinEffects)
         /// to the input feature. Used when the number of input features is different than the number of shape functions. Use default if all features have
         /// a shape function.</param>
-        public RegressionGamModelParameters(IHostEnvironment env,
+        internal RegressionGamModelParameters(IHostEnvironment env,
             double[][] binUpperBounds, double[][] binEffects, double intercept, int inputLength = -1, int[] featureToInputMap = null)
             : base(env, LoaderSignature, binUpperBounds, binEffects, intercept, inputLength, featureToInputMap) { }
 

src/Microsoft.ML.PCA/PcaTrainer.cs

@@ -411,7 +411,7 @@ private static VersionInfo GetVersionInfo()
         /// <param name="rank">The rank of the PCA approximation of the covariance matrix. This is the number of eigenvectors in the model.</param>
         /// <param name="eigenVectors">Array of eigenvectors.</param>
         /// <param name="mean">The mean vector of the training data.</param>
-        public PcaModelParameters(IHostEnvironment env, int rank, float[][] eigenVectors, in VBuffer<float> mean)
+        internal PcaModelParameters(IHostEnvironment env, int rank, float[][] eigenVectors, in VBuffer<float> mean)
             : base(env, RegistrationName)
         {
             _dimension = eigenVectors[0].Length;

src/Microsoft.ML.StandardLearners/FactorizationMachine/FieldAwareFactorizationMachineModelParameters.cs

@@ -57,7 +57,7 @@ private static VersionInfo GetVersionInfo()
         /// and each latent vector contains <see cref="LatentDim"/> values. In the f-th field, the j-th feature's latent vector, `v_{j, f}` in the doc
         /// https://github.com/wschin/fast-ffm/blob/master/fast-ffm.pdf, starts at latentWeights[j * fieldCount * latentDim + f * latentDim].
         /// The k-th element in v_{j, f} is latentWeights[j * fieldCount * latentDim + f * latentDim + k]. The size of the array must be featureCount x fieldCount x latentDim.</param>
-        public FieldAwareFactorizationMachineModelParameters(IHostEnvironment env, bool norm, int fieldCount, int featureCount, int latentDim,
+        internal FieldAwareFactorizationMachineModelParameters(IHostEnvironment env, bool norm, int fieldCount, int featureCount, int latentDim,
             float[] linearWeights, float[] latentWeights) : base(env, LoaderSignature)
         {
             Host.Assert(fieldCount > 0);

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

@@ -115,7 +115,7 @@ public IEnumerator<float> GetEnumerator()
         /// <param name="weights">The weights for the linear model. The i-th element of weights is the coefficient
         /// of the i-th feature. Note that this will take ownership of the <see cref="VBuffer{T}"/>.</param>
         /// <param name="bias">The bias added to every output score.</param>
-        public LinearModelParameters(IHostEnvironment env, string name, in VBuffer<float> weights, float bias)
+        internal LinearModelParameters(IHostEnvironment env, string name, in VBuffer<float> weights, float bias)
             : base(env, name)
         {
             Host.CheckParam(FloatUtils.IsFinite(weights.GetValues()), nameof(weights), "Cannot initialize linear predictor with non-finite weights");
@@ -436,7 +436,7 @@ private static VersionInfo GetVersionInfo()
         /// of the i-th feature. Note that this will take ownership of the <see cref="VBuffer{T}"/>.</param>
         /// <param name="bias">The bias added to every output score.</param>
         /// <param name="stats"></param>
-        public LinearBinaryModelParameters(IHostEnvironment env, in VBuffer<float> weights, float bias, LinearModelStatistics stats = null)
+        internal LinearBinaryModelParameters(IHostEnvironment env, in VBuffer<float> weights, float bias, LinearModelStatistics stats = null)
             : base(env, RegistrationName, in weights, bias)
         {
             Contracts.AssertValueOrNull(stats);
@@ -604,7 +604,7 @@ private static VersionInfo GetVersionInfo()
         /// <param name="weights">The weights for the linear model. The i-th element of weights is the coefficient
         /// of the i-th feature. Note that this will take ownership of the <see cref="VBuffer{T}"/>.</param>
         /// <param name="bias">The bias added to every output score.</param>
-        public LinearRegressionModelParameters(IHostEnvironment env, in VBuffer<float> weights, float bias)
+        internal LinearRegressionModelParameters(IHostEnvironment env, in VBuffer<float> weights, float bias)
             : base(env, RegistrationName, in weights, bias)
         {
         }
@@ -687,7 +687,7 @@ private static VersionInfo GetVersionInfo()
         /// <param name="weights">The weights for the linear model. The i-th element of weights is the coefficient
         /// of the i-th feature. Note that this will take ownership of the <see cref="VBuffer{T}"/>.</param>
         /// <param name="bias">The bias added to every output score.</param>
-        public PoissonRegressionModelParameters(IHostEnvironment env, in VBuffer<float> weights, float bias)
+        internal PoissonRegressionModelParameters(IHostEnvironment env, in VBuffer<float> weights, float bias)
             : base(env, RegistrationName, in weights, bias)
         {
         }

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

@@ -434,7 +434,7 @@ internal MulticlassLogisticRegressionModelParameters(IHostEnvironment env, in VB
         /// <param name="numFeatures">The length of the feature vector.</param>
         /// <param name="labelNames">The optional label names. If specified not null, it should have the same length as <paramref name="numClasses"/>.</param>
         /// <param name="stats">The model statistics.</param>
-        public MulticlassLogisticRegressionModelParameters(IHostEnvironment env, VBuffer<float>[] weights, float[] bias, int numClasses, int numFeatures, string[] labelNames, LinearModelStatistics stats = null)
+        internal MulticlassLogisticRegressionModelParameters(IHostEnvironment env, VBuffer<float>[] weights, float[] bias, int numClasses, int numFeatures, string[] labelNames, LinearModelStatistics stats = null)
             : base(env, RegistrationName)
         {
             Contracts.CheckValue(weights, nameof(weights));