Revert "Revert "Add type resolving for ONNX models""

This reverts commit 882d574ab53e8d10a2a8765a64487c20661dc63f.

24 files changed, 1293 insertions, 149 deletions

diff --git a/config-model/pom.xml b/config-model/pom.xml
index 95e79fd09fb..c0751431d03 100644
--- a/config-model/pom.xml
+++ b/config-model/pom.xml
@@ -46,6 +46,11 @@
       <scope>test</scope>
     </dependency>
     <dependency>
+      <groupId>com.google.protobuf</groupId>
+      <artifactId>protobuf-java</artifactId>
+      <version>${protobuf.version}</version>
+    </dependency>
+    <dependency>
       <groupId>com.google.guava</groupId>
       <artifactId>guava</artifactId>
       <scope>provided</scope>
@@ -498,6 +503,10 @@
           <updateReleaseInfo>true</updateReleaseInfo>
         </configuration>
       </plugin>
+      <plugin>
+        <groupId>com.github.os72</groupId>
+        <artifactId>protoc-jar-maven-plugin</artifactId>
+      </plugin>
     </plugins>
   </build>
 </project>
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/MapEvaluationTypeContext.java b/config-model/src/main/java/com/yahoo/searchdefinition/MapEvaluationTypeContext.java
index 4011ce43841..b153ff62e7d 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/MapEvaluationTypeContext.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/MapEvaluationTypeContext.java
@@ -1,6 +1,7 @@
 // Copyright 2018 Yahoo Holdings. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
 package com.yahoo.searchdefinition;
 
+import com.yahoo.searchdefinition.expressiontransforms.OnnxModelTransformer;
 import com.yahoo.searchlib.rankingexpression.ExpressionFunction;
 import com.yahoo.searchlib.rankingexpression.RankingExpression;
 import com.yahoo.searchlib.rankingexpression.Reference;
@@ -158,6 +159,12 @@ public class MapEvaluationTypeContext extends FunctionReferenceContext implement
                 return function.get().getBody().type(this.withBindings(bind(function.get().arguments(), reference.arguments())));
             }
 
+            // A reference to an ONNX model?
+            Optional<TensorType> onnxFeatureType = onnxFeatureType(reference);
+            if (onnxFeatureType.isPresent()) {
+                return onnxFeatureType.get();
+            }
+
             // A reference to a feature which returns a tensor?
             Optional<TensorType> featureTensorType = tensorFeatureType(reference);
             if (featureTensorType.isPresent()) {
@@ -210,6 +217,26 @@ public class MapEvaluationTypeContext extends FunctionReferenceContext implement
         return Optional.of(function);
     }
 
+    private Optional<TensorType> onnxFeatureType(Reference reference) {
+        if ( ! reference.name().equals("onnxModel"))
+            return Optional.empty();
+
+        if ( ! featureTypes.containsKey(reference)) {
+            String configOrFileName = reference.arguments().expressions().get(0).toString();
+
+            // Look up standardized format as added in RankProfile
+            String modelConfigName = OnnxModelTransformer.getModelConfigName(reference);
+            String modelOutput = OnnxModelTransformer.getModelOutput(reference, null);
+
+            reference  = new Reference("onnxModel", new Arguments(new ReferenceNode(modelConfigName)), modelOutput);
+            if ( ! featureTypes.containsKey(reference)) {
+                throw new IllegalArgumentException("Missing onnx-model config for '" + configOrFileName + "'");
+            }
+        }
+
+        return Optional.of(featureTypes.get(reference));
+    }
+
     /**
      * There are two features which returns the (non-empty) tensor type: tensorFromLabels and tensorFromWeightedSet.
      * This returns the type of those features if this is a reference to either of them, or empty otherwise.
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/OnnxModel.java b/config-model/src/main/java/com/yahoo/searchdefinition/OnnxModel.java
index c2fb2107604..58213186f78 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/OnnxModel.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/OnnxModel.java
@@ -2,14 +2,22 @@
 package com.yahoo.searchdefinition;
 
 import com.yahoo.config.FileReference;
+import com.yahoo.path.Path;
+import com.yahoo.searchlib.rankingexpression.ExpressionFunction;
+import com.yahoo.searchlib.rankingexpression.Reference;
+import com.yahoo.tensor.TensorType;
 import com.yahoo.vespa.model.AbstractService;
 import com.yahoo.vespa.model.utils.FileSender;
+import onnx.Onnx;
 
-import java.util.ArrayList;
 import java.util.Collection;
 import java.util.Collections;
-import java.util.List;
+import java.util.HashMap;
+import java.util.HashSet;
+import java.util.Map;
 import java.util.Objects;
+import java.util.Optional;
+import java.util.Set;
 
 /**
  * A global ONNX model distributed using file distribution, similar to ranking constants.
@@ -21,16 +29,16 @@ public class OnnxModel {
     public enum PathType {FILE, URI};
 
     private final String name;
+    private PathType pathType = PathType.FILE;
     private String path = null;
     private String fileReference = "";
-    private List<OnnxNameMapping> inputMap = new ArrayList<>();
-    private List<OnnxNameMapping> outputMap = new ArrayList<>();
+    private String defaultOutput = null;
+    private Map<String, String> inputMap = new HashMap<>();
+    private Map<String, String> outputMap = new HashMap<>();
 
-    public PathType getPathType() {
-        return pathType;
-    }
-
-    private PathType pathType = PathType.FILE;
+    private Map<String, Onnx.TypeProto> inputTypes = new HashMap<>();
+    private Map<String, Onnx.TypeProto> outputTypes = new HashMap<>();
+    private Map<String, TensorType>     vespaTypes = new HashMap<>();
 
     public OnnxModel(String name) {
         this.name = name;
@@ -49,21 +57,52 @@ public class OnnxModel {
     }
 
     public void setUri(String uri) {
-        Objects.requireNonNull(uri, "uri cannot be null");
-        this.path = uri;
-        this.pathType = PathType.URI;
+        throw new IllegalArgumentException("URI for ONNX models are not currently supported");
+    }
+
+    public PathType getPathType() {
+        return pathType;
+    }
+
+    public void setDefaultOutput(String onnxName) {
+        Objects.requireNonNull(onnxName, "Name cannot be null");
+        this.defaultOutput = onnxName;
     }
 
     public void addInputNameMapping(String onnxName, String vespaName) {
+        addInputNameMapping(onnxName, vespaName, true);
+    }
+
+    public void addInputNameMapping(String onnxName, String vespaName, boolean overwrite) {
         Objects.requireNonNull(onnxName, "Onnx name cannot be null");
         Objects.requireNonNull(vespaName, "Vespa name cannot be null");
-        this.inputMap.add(new OnnxNameMapping(onnxName, vespaName));
+        if (overwrite || ! inputMap.containsKey(onnxName)) {
+            inputMap.put(onnxName, vespaName);
+        }
     }
 
     public void addOutputNameMapping(String onnxName, String vespaName) {
+        addOutputNameMapping(onnxName, vespaName, true);
+    }
+
+    public void addOutputNameMapping(String onnxName, String vespaName, boolean overwrite) {
         Objects.requireNonNull(onnxName, "Onnx name cannot be null");
         Objects.requireNonNull(vespaName, "Vespa name cannot be null");
-        this.outputMap.add(new OnnxNameMapping(onnxName, vespaName));
+        if (overwrite || ! outputMap.containsKey(onnxName)) {
+            outputMap.put(onnxName, vespaName);
+        }
+    }
+
+    public void addInputType(String onnxName, Onnx.TypeProto type) {
+        Objects.requireNonNull(onnxName, "Onnx name cannot be null");
+        Objects.requireNonNull(type, "Tensor type cannot be null");
+        inputTypes.put(onnxName, type);
+    }
+
+    public void addOutputType(String onnxName, Onnx.TypeProto type) {
+        Objects.requireNonNull(onnxName, "Onnx name cannot be null");
+        Objects.requireNonNull(type, "Tensor type cannot be null");
+        outputTypes.put(onnxName, type);
     }
 
     /** Initiate sending of this constant to some services over file distribution */
@@ -76,11 +115,16 @@ public class OnnxModel {
 
     public String getName() { return name; }
     public String getFileName() { return path; }
+    public Path   getFilePath() { return Path.fromString(path); }
     public String getUri() { return path; }
     public String getFileReference() { return fileReference; }
 
-    public List<OnnxNameMapping> getInputMap() { return Collections.unmodifiableList(inputMap); }
-    public List<OnnxNameMapping> getOutputMap() { return Collections.unmodifiableList(outputMap); }
+    public Map<String, String> getInputMap() { return Collections.unmodifiableMap(inputMap); }
+    public Map<String, String> getOutputMap() { return Collections.unmodifiableMap(outputMap); }
+
+    public String getDefaultOutput() {
+        return defaultOutput;
+    }
 
     public void validate() {
         if (path == null || path.isEmpty())
@@ -90,23 +134,142 @@ public class OnnxModel {
     public String toString() {
         StringBuilder b = new StringBuilder();
         b.append("onnx-model '").append(name)
-         .append(pathType == PathType.FILE ? "' from file '" : " from uri ").append(path)
-         .append("' with ref '").append(fileReference)
-         .append("'");
+                .append(pathType == PathType.FILE ? "' from file '" : " from uri ").append(path)
+                .append("' with ref '").append(fileReference)
+                .append("'");
         return b.toString();
     }
 
-    public static class OnnxNameMapping {
-        private String onnxName;
-        private String vespaName;
+    /**
+     * Return the tensor type for an ONNX model for the given context.
+     * An ONNX model can have dynamic/symbolic dimension sizes. If so, the output
+     * type depends on the input types for the given context (rank profile).
+     */
+    public TensorType getTensorType(String onnxName, MapEvaluationTypeContext context) {
+        Onnx.TypeProto onnxOutputType = outputTypes.get(onnxName);
+        if (onnxOutputType == null) {
+            throw new IllegalArgumentException("Could not find type for output '" + onnxName + "' " + "in '" + name + "'");
+        }
+        if (containsSymbolicDimensionSizes(onnxOutputType)) {
+            return getTensorTypeWithSymbolicDimensions(onnxOutputType, context);
+        }
+        return vespaTypes.computeIfAbsent(onnxName, v -> typeFrom(onnxOutputType));
+    }
+
+    private TensorType getTensorTypeWithSymbolicDimensions(Onnx.TypeProto onnxOutputType, MapEvaluationTypeContext context) {
+        Map<String, Long> symbolicSizes = resolveSymbolicDimensionSizes(context);
+        if (symbolicSizes.isEmpty()) {
+            return TensorType.empty;  // Context is probably a rank profile not using this ONNX model
+        }
+        return typeFrom(onnxOutputType, symbolicSizes);
+    }
+
+    private Map<String, Long> resolveSymbolicDimensionSizes(MapEvaluationTypeContext context) {
+        Map<String, Long> symbolicSizes = new HashMap<>();
+        for (String onnxInputName : inputTypes.keySet()) {
+
+            Onnx.TypeProto onnxType = inputTypes.get(onnxInputName);
+            if ( ! containsSymbolicDimensionSizes(onnxType)) {
+                continue;
+            }
+
+            Optional<TensorType> vespaType = resolveInputType(onnxInputName, context);
+            if (vespaType.isEmpty()) {
+                return Collections.emptyMap();
+            }
+
+            var onnxDimensions = onnxType.getTensorType().getShape().getDimList();
+            var vespaDimensions = vespaType.get().dimensions();
+            if (vespaDimensions.size() != onnxDimensions.size()) {
+                return Collections.emptyMap();
+            }
+
+            for (int i = 0; i < vespaDimensions.size(); ++i) {
+                if (vespaDimensions.get(i).size().isEmpty() || ! onnxDimensions.get(i).hasDimParam()) {
+                    continue;
+                }
+                String symbolicName = onnxDimensions.get(i).getDimParam();
+                Long size = vespaDimensions.get(i).size().get();
+                if (symbolicSizes.containsKey(symbolicName) && ! symbolicSizes.get(symbolicName).equals(size)) {
+                    throw new IllegalArgumentException("Found conflicting sizes for symbolic dimension " +
+                            "'" + symbolicName + "' for input '" + onnxInputName + "' in ONNX model '" +  name + "'");
+                }
+                symbolicSizes.put(symbolicName, size);
+            }
+        }
+        return symbolicSizes;
+    }
+
+    private Optional<TensorType> resolveInputType(String onnxInputName, MapEvaluationTypeContext context) {
+        String source = inputMap.get(onnxInputName);
+        if (source != null) {
+            // Source is either a simple reference (query/attribute/constant)...
+            Optional<Reference> reference = Reference.simple(source);
+            if (reference.isPresent()) {
+                return Optional.of(context.getType(reference.get()));
+            }
+            // ... or a function
+            ExpressionFunction func = context.getFunction(source);
+            if (func != null) {
+                return Optional.of(func.getBody().type(context));
+            }
+        }
+        return Optional.empty();  // if this context does not contain this input
+    }
+
+    private static boolean containsSymbolicDimensionSizes(Onnx.TypeProto type) {
+        return type.getTensorType().getShape().getDimList().stream().anyMatch(d -> d.hasDimParam() && ! d.hasDimValue());
+    }
+
+    private static TensorType typeFrom(Onnx.TypeProto type) {
+        return typeFrom(type, null);
+    }
+
+    private static TensorType typeFrom(Onnx.TypeProto type, Map<String, Long> symbolicSizes) {
+        String dimensionPrefix = "d"; // standard naming convention: d0, d1, ...
+        Onnx.TensorShapeProto shape = type.getTensorType().getShape();
+        TensorType.Builder builder = new TensorType.Builder(toValueType(type.getTensorType().getElemType()));
+        for (int i = 0; i < shape.getDimCount(); ++ i) {
+            String dimensionName = dimensionPrefix + i;
+            Onnx.TensorShapeProto.Dimension onnxDimension = shape.getDim(i);
+            long onnxDimensionSize = onnxDimension.getDimValue();
+            if (onnxDimension.hasDimParam() && symbolicSizes != null && symbolicSizes.containsKey(onnxDimension.getDimParam())) {
+                onnxDimensionSize = symbolicSizes.get(onnxDimension.getDimParam());
+            }
+            if (onnxDimensionSize == 0 && symbolicSizes != null) {
+                // This is for the case where all symbolic dimensions have
+                // different names, but can be resolved to a single dimension size.
+                Set<Long> unknownSizes = new HashSet<>(symbolicSizes.values());
+                if (unknownSizes.size() == 1) {
+                    onnxDimensionSize = unknownSizes.iterator().next();
+                }
+            }
+            if (onnxDimensionSize <= 0) {
+                throw new IllegalArgumentException("Unable to determine fixed dimension size when converting from " +
+                        "ONNX type: " + type + " to Vespa tensor type.");
+            }
+            builder.indexed(dimensionName, onnxDimensionSize);
+        }
+        return builder.build();
+    }
 
-        private OnnxNameMapping(String onnxName, String vespaName) {
-            this.onnxName = onnxName;
-            this.vespaName = vespaName;
+    private static TensorType.Value toValueType(Onnx.TensorProto.DataType dataType) {
+        switch (dataType) {
+            case FLOAT: return TensorType.Value.FLOAT;
+            case DOUBLE: return TensorType.Value.DOUBLE;
+            // Imperfect conversion, for now:
+            case BOOL: return TensorType.Value.FLOAT;
+            case INT8: return TensorType.Value.FLOAT;
+            case INT16: return TensorType.Value.FLOAT;
+            case INT32: return TensorType.Value.FLOAT;
+            case INT64: return TensorType.Value.FLOAT;
+            case UINT8: return TensorType.Value.FLOAT;
+            case UINT16: return TensorType.Value.FLOAT;
+            case UINT32: return TensorType.Value.FLOAT;
+            case UINT64: return TensorType.Value.FLOAT;
+            default: throw new IllegalArgumentException("A ONNX tensor with data type " + dataType +
+                    " cannot be converted to a Vespa tensor type");
         }
-        public String getOnnxName() { return onnxName; }
-        public String getVespaName() { return vespaName; }
-        public void setVespaName(String vespaName) { this.vespaName = vespaName; }
     }
 
 }
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/RankProfile.java b/config-model/src/main/java/com/yahoo/searchdefinition/RankProfile.java
index d309f48d6df..96c043bdb34 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/RankProfile.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/RankProfile.java
@@ -18,6 +18,7 @@ import com.yahoo.searchlib.rankingexpression.RankingExpression;
 import com.yahoo.searchlib.rankingexpression.Reference;
 import com.yahoo.searchlib.rankingexpression.evaluation.TensorValue;
 import com.yahoo.searchlib.rankingexpression.evaluation.Value;
+import com.yahoo.searchlib.rankingexpression.rule.Arguments;
 import com.yahoo.searchlib.rankingexpression.rule.ReferenceNode;
 import com.yahoo.tensor.TensorType;
 import com.yahoo.vespa.model.VespaModel;
@@ -158,6 +159,10 @@ public class RankProfile implements Cloneable {
         return search != null ? search.rankingConstants() : model.rankingConstants();
     }
 
+    private Map<String, OnnxModel> onnxModels() {
+        return search != null ? search.onnxModels().asMap() : Collections.emptyMap();
+    }
+
     private Stream<ImmutableSDField> allFields() {
         if (search == null) return Stream.empty();
         if (allFieldsList == null) {
@@ -821,6 +826,20 @@ public class RankProfile implements Cloneable {
             }
         }
 
+        // Add output types for ONNX models
+        for (Map.Entry<String, OnnxModel> entry : onnxModels().entrySet()) {
+            String modelName = entry.getKey();
+            OnnxModel model = entry.getValue();
+            Arguments args = new Arguments(new ReferenceNode(modelName));
+
+            TensorType defaultOutputType = model.getTensorType(model.getDefaultOutput(), context);
+            context.setType(new Reference("onnxModel", args, null), defaultOutputType);
+
+            for (Map.Entry<String, String> mapping : model.getOutputMap().entrySet()) {
+                TensorType type = model.getTensorType(mapping.getKey(), context);
+                context.setType(new Reference("onnxModel", args, mapping.getValue()), type);
+            }
+        }
         return context;
     }
 
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/derived/RankProfileList.java b/config-model/src/main/java/com/yahoo/searchdefinition/derived/RankProfileList.java
index 84442fedc48..22a32c8fd65 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/derived/RankProfileList.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/derived/RankProfileList.java
@@ -126,8 +126,8 @@ public class RankProfileList extends Derived implements RankProfilesConfig.Produ
                 OnnxModelsConfig.Model.Builder modelBuilder = new OnnxModelsConfig.Model.Builder();
                 modelBuilder.name(model.getName());
                 modelBuilder.fileref(model.getFileReference());
-                model.getInputMap().forEach(mapper -> modelBuilder.input(new OnnxModelsConfig.Model.Input.Builder().name(mapper.getOnnxName()).source(mapper.getVespaName())));
-                model.getOutputMap().forEach(mapper -> modelBuilder.output(new OnnxModelsConfig.Model.Output.Builder().name(mapper.getOnnxName()).as(mapper.getVespaName())));
+                model.getInputMap().forEach((name, source) -> modelBuilder.input(new OnnxModelsConfig.Model.Input.Builder().name(name).source(source)));
+                model.getOutputMap().forEach((name, as) -> modelBuilder.output(new OnnxModelsConfig.Model.Output.Builder().name(name).as(as)));
                 builder.model(modelBuilder);
             }
         }
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/derived/RawRankProfile.java b/config-model/src/main/java/com/yahoo/searchdefinition/derived/RawRankProfile.java
index 87eaaf0387a..56a5d539906 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/derived/RawRankProfile.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/derived/RawRankProfile.java
@@ -448,10 +448,10 @@ public class RawRankProfile implements RankProfilesConfig.Producer {
             Set<String> functionNames = rankProfile.getFunctions().keySet();
             if (functionNames.isEmpty()) return;
             for (OnnxModel onnxModel: rankProfile.getSearch().onnxModels().asMap().values()) {
-                for (OnnxModel.OnnxNameMapping mapping : onnxModel.getInputMap()) {
-                    String source = mapping.getVespaName();
+                for (Map.Entry<String, String> mapping : onnxModel.getInputMap().entrySet()) {
+                    String source = mapping.getValue();
                     if (functionNames.contains(source)) {
-                        mapping.setVespaName("rankingExpression(" + source + ")");
+                        onnxModel.addInputNameMapping(mapping.getKey(), "rankingExpression(" + source + ")");
                     }
                 }
             }
@@ -462,7 +462,7 @@ public class RawRankProfile implements RankProfilesConfig.Producer {
             Set<ReferenceNode> replacedSummaryFeatures = new HashSet<>();
             for (Iterator<ReferenceNode> i = summaryFeatures.iterator(); i.hasNext(); ) {
                 ReferenceNode referenceNode = i.next();
-                ReferenceNode replacedNode = OnnxModelTransformer.transformFeature(referenceNode, rankProfile.getSearch());
+                ReferenceNode replacedNode = (ReferenceNode) OnnxModelTransformer.transformFeature(referenceNode, rankProfile);
                 if (referenceNode != replacedNode) {
                     replacedSummaryFeatures.add(replacedNode);
                     i.remove();
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/OnnxFeatureConverter.java b/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/OnnxFeatureConverter.java
index ec517768ea9..d23a8376e7a 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/OnnxFeatureConverter.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/OnnxFeatureConverter.java
@@ -38,7 +38,7 @@ public class OnnxFeatureConverter extends ExpressionTransformer<RankProfileTrans
     }
 
     private ExpressionNode transformFeature(ReferenceNode feature, RankProfileTransformContext context) {
-        if ( ! feature.getName().equals("onnx")) return feature;
+        if ( ! feature.getName().equals("onnx") && ! feature.getName().equals("onnx_vespa")) return feature;
 
         try {
             FeatureArguments arguments = asFeatureArguments(feature.getArguments());
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/OnnxModelTransformer.java b/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/OnnxModelTransformer.java
index e1ad003e5bd..69cdae10e47 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/OnnxModelTransformer.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/OnnxModelTransformer.java
@@ -1,20 +1,36 @@
 // Copyright Verizon Media. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
 package com.yahoo.searchdefinition.expressiontransforms;
 
+import com.yahoo.path.Path;
 import com.yahoo.searchdefinition.ImmutableSearch;
 import com.yahoo.searchdefinition.OnnxModel;
+import com.yahoo.searchdefinition.RankProfile;
+import com.yahoo.searchlib.rankingexpression.Reference;
 import com.yahoo.searchlib.rankingexpression.rule.Arguments;
 import com.yahoo.searchlib.rankingexpression.rule.CompositeNode;
 import com.yahoo.searchlib.rankingexpression.rule.ConstantNode;
 import com.yahoo.searchlib.rankingexpression.rule.ExpressionNode;
 import com.yahoo.searchlib.rankingexpression.rule.ReferenceNode;
 import com.yahoo.searchlib.rankingexpression.transform.ExpressionTransformer;
+import com.yahoo.vespa.model.ml.ConvertedModel;
+import com.yahoo.vespa.model.ml.FeatureArguments;
+import com.yahoo.vespa.model.ml.ModelName;
 
 import java.util.List;
 
 /**
- * Transforms instances of the onnxModel ranking feature and generates
- * ONNX configuration if necessary.
+ * Transforms ONNX model features of the forms:
+ *
+ *     onnxModel(config_name)
+ *     onnxModel(config_name).output
+ *     onnxModel("path/to/model")
+ *     onnxModel("path/to/model").output
+ *     onnxModel("path/to/model", "path/to/output")
+ *     onnxModel("path/to/model", "unused", "path/to/output")    // signature is unused
+ *
+ * To the format expected by the backend:
+ *
+ *     onnxModel(config_name).output
  *
  * @author lesters
  */
@@ -33,85 +49,92 @@ public class OnnxModelTransformer extends ExpressionTransformer<RankProfileTrans
     private ExpressionNode transformFeature(ReferenceNode feature, RankProfileTransformContext context) {
         if (context.rankProfile() == null) return feature;
         if (context.rankProfile().getSearch() == null) return feature;
-        return transformFeature(feature, context.rankProfile().getSearch());
+        return transformFeature(feature, context.rankProfile());
     }
 
-    public static ReferenceNode transformFeature(ReferenceNode feature, ImmutableSearch search) {
-        if (!feature.getName().equals("onnxModel")) return feature;
+    public static ExpressionNode transformFeature(ReferenceNode feature, RankProfile rankProfile) {
+        ImmutableSearch search = rankProfile.getSearch();
+        final String featureName = feature.getName();
+        if ( ! featureName.equals("onnxModel")) return feature;
 
         Arguments arguments = feature.getArguments();
         if (arguments.isEmpty())
-            throw new IllegalArgumentException("An onnxModel feature must take an argument referring to a " +
-                    "onnx-model config or a ONNX file.");
-        if (arguments.expressions().size() > 2)
-            throw new IllegalArgumentException("An onnxModel feature can have at most 2 arguments.");
-
-        // Validation that the file actually exists is handled when the file is added to file distribution.
-        // Validation of inputs, outputs and corresponding types are currently handled by RankSetupValidator.
-
-        String modelConfigName;
-        OnnxModel onnxModel;
-        if (arguments.expressions().get(0) instanceof ReferenceNode) {
-            modelConfigName = arguments.expressions().get(0).toString();
-            onnxModel = search.onnxModels().get(modelConfigName);
-            if (onnxModel == null) {
-                throw new IllegalArgumentException("onnxModel argument '" + modelConfigName + "' config not found");
-            }
-        } else if (arguments.expressions().get(0) instanceof ConstantNode) {
+            throw new IllegalArgumentException("An " + featureName + " feature must take an argument referring to a " +
+                    "onnx-model config or an ONNX file.");
+        if (arguments.expressions().size() > 3)
+            throw new IllegalArgumentException("An " + featureName + " feature can have at most 3 arguments.");
+
+        // Check that the model configuration "onnx-model" exists. If not defined, it should have been added
+        // by the "OnnxModelConfigGenerator" processor. If it still doesn't exist, it is because we can't find
+        // the actual ONNX file, which can happen if we are restarting or upgrading an application using an
+        // ONNX file that was transformed to Vespa ranking expressions. We then assume it is in the model store.
+
+        String modelConfigName = getModelConfigName(feature.reference());
+        OnnxModel onnxModel = search.onnxModels().get(modelConfigName);
+        if (onnxModel == null) {
             String path = asString(arguments.expressions().get(0));
-            modelConfigName = asValidIdentifier(path);
-            onnxModel = search.onnxModels().get(modelConfigName);
-            if (onnxModel == null) {
-                onnxModel = new OnnxModel(modelConfigName, path);
-                search.onnxModels().add(onnxModel);
-            }
-        } else {
-            throw new IllegalArgumentException("Illegal argument to onnxModel: '" + arguments.expressions().get(0) + "'");
+            ModelName modelName = new ModelName(null, Path.fromString(path), true);
+            ConvertedModel convertedModel = ConvertedModel.fromStore(modelName, path, rankProfile);
+            FeatureArguments featureArguments = new FeatureArguments(arguments);
+            return convertedModel.expression(featureArguments, null);
         }
 
-        String output = null;
-        if (feature.getOutput() != null) {
-            output = feature.getOutput();
-            if ( ! hasOutputMapping(onnxModel, output)) {
-                onnxModel.addOutputNameMapping(output, output);
+        String defaultOutput = onnxModel.getOutputMap().get(onnxModel.getDefaultOutput());
+        String output = getModelOutput(feature.reference(), defaultOutput);
+        if (! onnxModel.getOutputMap().containsValue(output)) {
+            throw new IllegalArgumentException(featureName + " argument '" + output +
+                    "' output not found in model '" + onnxModel.getFileName() + "'");
+        }
+        return new ReferenceNode("onnxModel", List.of(new ReferenceNode(modelConfigName)), output);
+    }
+
+    public static String getModelConfigName(Reference reference) {
+        if (reference.arguments().size() > 0) {
+            ExpressionNode expr = reference.arguments().expressions().get(0);
+            if (expr instanceof ReferenceNode) {  // refers to onnx-model config
+                return expr.toString();
             }
-        } else if (arguments.expressions().size() > 1) {
-            String name = asString(arguments.expressions().get(1));
-            output = asValidIdentifier(name);
-            if ( ! hasOutputMapping(onnxModel, output)) {
-                onnxModel.addOutputNameMapping(name, output);
+            if (expr instanceof ConstantNode) {  // refers to an file path
+                return asValidIdentifier(expr);
             }
         }
+        return null;
+    }
 
-        // Replace feature with name of config
-        ExpressionNode argument = new ReferenceNode(modelConfigName);
-        return new ReferenceNode("onnxModel", List.of(argument), output);
-
+    public static String getModelOutput(Reference reference, String defaultOutput) {
+        if (reference.output() != null) {
+            return reference.output();
+        } else if (reference.arguments().expressions().size() == 2) {
+            return asValidIdentifier(reference.arguments().expressions().get(1));
+        } else if (reference.arguments().expressions().size() > 2) {
+            return asValidIdentifier(reference.arguments().expressions().get(2));
+        }
+        return defaultOutput;
     }
 
-    private static boolean hasOutputMapping(OnnxModel onnxModel, String as) {
-        return onnxModel.getOutputMap().stream().anyMatch(m -> m.getVespaName().equals(as));
+    public static String stripQuotes(String s) {
+        if (isNotQuoteSign(s.codePointAt(0))) return s;
+        if (isNotQuoteSign(s.codePointAt(s.length() - 1)))
+            throw new IllegalArgumentException("argument [" + s + "] is missing end quote");
+        return s.substring(1, s.length()-1);
     }
 
-    private static String asString(ExpressionNode node) {
-        if ( ! (node instanceof ConstantNode))
-            throw new IllegalArgumentException("Expected a constant string as argument, but got '" + node);
-        return stripQuotes(((ConstantNode)node).sourceString());
+    public static String asValidIdentifier(String str) {
+        return str.replaceAll("[^\\w\\d\\$@_]", "_");
     }
 
-    private static String stripQuotes(String s) {
-        if ( ! isQuoteSign(s.codePointAt(0))) return s;
-        if ( ! isQuoteSign(s.codePointAt(s.length() - 1 )))
-            throw new IllegalArgumentException("argument [" + s + "] is missing endquote");
-        return s.substring(1, s.length()-1);
+    private static String asValidIdentifier(ExpressionNode node) {
+        return asValidIdentifier(asString(node));
     }
 
-    private static boolean isQuoteSign(int c) {
-        return c == '\'' || c == '"';
+    private static boolean isNotQuoteSign(int c) {
+        return c != '\'' && c != '"';
     }
 
-    private static String asValidIdentifier(String str) {
-        return str.replaceAll("[^\\w\\d\\$@_]", "_");
+    public static String asString(ExpressionNode node) {
+        if ( ! (node instanceof ConstantNode))
+            throw new IllegalArgumentException("Expected a constant string as argument, but got '" + node);
+        return stripQuotes(((ConstantNode)node).sourceString());
     }
 
 }
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/processing/OnnxModelConfigGenerator.java b/config-model/src/main/java/com/yahoo/searchdefinition/processing/OnnxModelConfigGenerator.java
new file mode 100644
index 00000000000..afba88c135d
--- /dev/null
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/processing/OnnxModelConfigGenerator.java
@@ -0,0 +1,97 @@
+// Copyright Verizon Media. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+
+package com.yahoo.searchdefinition.processing;
+
+import com.yahoo.config.application.api.ApplicationPackage;
+import com.yahoo.config.application.api.DeployLogger;
+import com.yahoo.searchdefinition.OnnxModel;
+import com.yahoo.searchdefinition.RankProfile;
+import com.yahoo.searchdefinition.RankProfileRegistry;
+import com.yahoo.searchdefinition.Search;
+import com.yahoo.searchdefinition.expressiontransforms.OnnxModelTransformer;
+import com.yahoo.searchlib.rankingexpression.rule.CompositeNode;
+import com.yahoo.searchlib.rankingexpression.rule.ConstantNode;
+import com.yahoo.searchlib.rankingexpression.rule.ExpressionNode;
+import com.yahoo.searchlib.rankingexpression.rule.ReferenceNode;
+import com.yahoo.vespa.model.container.search.QueryProfiles;
+
+import java.util.Map;
+
+/**
+ * Processes ONNX ranking features of the form:
+ *
+ *     onnx("files/model.onnx", "path/to/output:1")
+ *
+ * And generates an "onnx-model" configuration as if it was defined in the schema:
+ *
+ *   onnx-model files_model_onnx {
+ *       file:  "files/model.onnx"
+ *   }
+ *
+ * Inputs and outputs are resolved in OnnxModelTypeResolver, which must be
+ * processed after this.
+ *
+ * @author lesters
+ */
+public class OnnxModelConfigGenerator extends Processor {
+
+    public OnnxModelConfigGenerator(Search search, DeployLogger deployLogger, RankProfileRegistry rankProfileRegistry, QueryProfiles queryProfiles) {
+        super(search, deployLogger, rankProfileRegistry, queryProfiles);
+    }
+
+    @Override
+    public void process(boolean validate, boolean documentsOnly) {
+        if (documentsOnly) return;
+        for (RankProfile profile : rankProfileRegistry.rankProfilesOf(search)) {
+            if (profile.getFirstPhaseRanking() != null) {
+                process(profile.getFirstPhaseRanking().getRoot());
+            }
+            if (profile.getSecondPhaseRanking() != null) {
+                process(profile.getSecondPhaseRanking().getRoot());
+            }
+            for (Map.Entry<String, RankProfile.RankingExpressionFunction> function : profile.getFunctions().entrySet()) {
+                process(function.getValue().function().getBody().getRoot());
+            }
+            for (ReferenceNode feature : profile.getSummaryFeatures()) {
+                process(feature);
+            }
+        }
+    }
+
+    private void process(ExpressionNode node) {
+        if (node instanceof ReferenceNode) {
+            process((ReferenceNode)node);
+        } else if (node instanceof CompositeNode) {
+            for (ExpressionNode child : ((CompositeNode) node).children()) {
+                process(child);
+            }
+        }
+    }
+
+    private void process(ReferenceNode feature) {
+        if (feature.getName().equals("onnxModel") || feature.getName().equals("onnx")) {
+            if (feature.getArguments().size() > 0) {
+                if (feature.getArguments().expressions().get(0) instanceof ConstantNode) {
+                    ConstantNode node = (ConstantNode) feature.getArguments().expressions().get(0);
+                    String path = OnnxModelTransformer.stripQuotes(node.sourceString());
+                    String modelConfigName = OnnxModelTransformer.asValidIdentifier(path);
+
+                    // Only add the configuration if the model can actually be found.
+                    if ( ! OnnxModelTypeResolver.modelFileExists(path, search.applicationPackage())) {
+                        path = ApplicationPackage.MODELS_DIR.append(path).toString();
+                        if ( ! OnnxModelTypeResolver.modelFileExists(path, search.applicationPackage())) {
+                            return;
+                        }
+                    }
+
+                    OnnxModel onnxModel = search.onnxModels().get(modelConfigName);
+                    if (onnxModel == null) {
+                        onnxModel = new OnnxModel(modelConfigName, path);
+                        search.onnxModels().add(onnxModel);
+                    }
+                }
+            }
+        }
+    }
+
+}
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/processing/OnnxModelTypeResolver.java b/config-model/src/main/java/com/yahoo/searchdefinition/processing/OnnxModelTypeResolver.java
new file mode 100644
index 00000000000..bead2e7e7c9
--- /dev/null
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/processing/OnnxModelTypeResolver.java
@@ -0,0 +1,154 @@
+// Copyright Verizon Media. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+
+package com.yahoo.searchdefinition.processing;
+
+import com.yahoo.cloud.config.ConfigserverConfig;
+import com.yahoo.component.Version;
+import com.yahoo.config.FileReference;
+import com.yahoo.config.application.api.ApplicationFile;
+import com.yahoo.config.application.api.ApplicationPackage;
+import com.yahoo.config.application.api.DeployLogger;
+import com.yahoo.config.application.api.FileRegistry;
+import com.yahoo.path.Path;
+import com.yahoo.searchdefinition.OnnxModel;
+import com.yahoo.searchdefinition.RankProfileRegistry;
+import com.yahoo.searchdefinition.Search;
+import com.yahoo.searchdefinition.expressiontransforms.OnnxModelTransformer;
+import com.yahoo.vespa.defaults.Defaults;
+import com.yahoo.vespa.model.container.search.QueryProfiles;
+import onnx.Onnx;
+
+import java.io.File;
+import java.io.FileInputStream;
+import java.io.FileNotFoundException;
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.file.Paths;
+import java.util.Map;
+import java.util.Optional;
+
+/**
+ * Processes every "onnx-model" element in the schema. Parses the model file,
+ * adds missing input and output mappings (assigning default names), and
+ * adds tensor types to all model inputs and outputs.
+ *
+ * Must be processed before RankingExpressingTypeResolver.
+ *
+ * @author lesters
+ */
+public class OnnxModelTypeResolver extends Processor {
+
+    public OnnxModelTypeResolver(Search search, DeployLogger deployLogger, RankProfileRegistry rankProfileRegistry, QueryProfiles queryProfiles) {
+        super(search, deployLogger, rankProfileRegistry, queryProfiles);
+    }
+
+    @Override
+    public void process(boolean validate, boolean documentsOnly) {
+        if (documentsOnly) return;
+
+        for (Map.Entry<String, OnnxModel> entry : search.onnxModels().asMap().entrySet())  {
+            OnnxModel modelConfig = entry.getValue();
+            try (InputStream inputStream = openModelFile(modelConfig.getFilePath())) {
+                Onnx.ModelProto model = Onnx.ModelProto.parseFrom(inputStream);
+
+                // Model inputs - if not defined, assumes a function is provided with a valid name
+                for (Onnx.ValueInfoProto valueInfo : model.getGraph().getInputList()) {
+                    String onnxInputName = valueInfo.getName();
+                    String vespaInputName = OnnxModelTransformer.asValidIdentifier(onnxInputName);
+                    modelConfig.addInputNameMapping(onnxInputName, vespaInputName, false);
+                    modelConfig.addInputType(onnxInputName, valueInfo.getType());
+                }
+
+                // Model outputs
+                for (Onnx.ValueInfoProto valueInfo : model.getGraph().getOutputList()) {
+                    String onnxOutputName = valueInfo.getName();
+                    String vespaOutputName = OnnxModelTransformer.asValidIdentifier(onnxOutputName);
+                    modelConfig.addOutputNameMapping(onnxOutputName, vespaOutputName, false);
+                    modelConfig.addOutputType(onnxOutputName, valueInfo.getType());
+                }
+
+                // Set the first output as default
+                if ( ! model.getGraph().getOutputList().isEmpty()) {
+                    modelConfig.setDefaultOutput(model.getGraph().getOutput(0).getName());
+                }
+
+            } catch (IOException e) {
+                throw new IllegalArgumentException("Unable to parse ONNX model", e);
+            }
+        }
+    }
+
+    static boolean modelFileExists(String path, ApplicationPackage app) {
+        Path pathInApplicationPackage = Path.fromString(path);
+        if (getFile(pathInApplicationPackage, app).exists()) {
+            return true;
+        }
+        if (getFileReference(pathInApplicationPackage, app).isPresent()) {
+            return true;
+        }
+        return false;
+    }
+
+    private InputStream openModelFile(Path path) throws FileNotFoundException {
+        ApplicationFile file;
+        Optional<FileReference> reference;
+        Path modelsPath = ApplicationPackage.MODELS_DIR.append(path);
+
+        if ((file = getFile(path)).exists()) {
+            return file.createInputStream();
+        }
+        if ((file = getFile(modelsPath)).exists()) {
+            return file.createInputStream();
+        }
+        if ((reference = getFileReference(path)).isPresent()) {
+            return openFromFileRepository(path, reference.get());
+        }
+        if ((reference = getFileReference(modelsPath)).isPresent()) {
+            return openFromFileRepository(modelsPath, reference.get());
+        }
+
+        throw new IllegalArgumentException("Unable to find ONNX model file \"" + path + "\" " +
+            "in application package or file repository.");
+    }
+
+    private ApplicationFile getFile(Path path) {
+        return getFile(path, search.applicationPackage());
+    }
+
+    private static ApplicationFile getFile(Path path, ApplicationPackage app) {
+        return app.getFile(path);
+    }
+
+    private static InputStream openFromFileRepository(Path path, FileReference reference) throws FileNotFoundException {
+        return new FileInputStream(new File(getFileRepositoryPath(path, reference.value())));
+    }
+
+    public static String getFileRepositoryPath(Path path, String fileReference) {
+        ConfigserverConfig cfg = new ConfigserverConfig(new ConfigserverConfig.Builder());  // assume defaults
+        String fileRefDir = Defaults.getDefaults().underVespaHome(cfg.fileReferencesDir());
+        return Paths.get(fileRefDir, fileReference, path.getName()).toString();
+    }
+
+    private Optional<FileReference> getFileReference(Path path) {
+        return getFileReference(path, search.applicationPackage());
+    }
+
+    private static Optional<FileReference> getFileReference(Path path, ApplicationPackage app) {
+        Optional<FileRegistry> fileRegistry = getLatestFileRegistry(app);
+        if (fileRegistry.isPresent()) {
+            for (FileRegistry.Entry file : fileRegistry.get().export()) {
+                if (file.relativePath.equals(path.toString())) {
+                    return Optional.of(file.reference);
+                }
+            }
+        }
+        return Optional.empty();
+    }
+
+    private static Optional<FileRegistry> getLatestFileRegistry(ApplicationPackage app) {
+        if (app == null) return Optional.empty();
+        Optional<Version> latest = app.getFileRegistries().keySet().stream().max(Version::compareTo);
+        return latest.isEmpty() ? Optional.empty() : Optional.of(app.getFileRegistries().get(latest.get()));
+    }
+
+}
diff --git a/config-model/src/main/java/com/yahoo/searchdefinition/processing/Processing.java b/config-model/src/main/java/com/yahoo/searchdefinition/processing/Processing.java
index e8594c2a87f..1a3ef9e54b4 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/processing/Processing.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/processing/Processing.java
@@ -74,6 +74,8 @@ public class Processing {
                 ReferenceFieldsProcessor::new,
                 FastAccessValidator::new,
                 ReservedFunctionNames::new,
+                OnnxModelConfigGenerator::new,
+                OnnxModelTypeResolver::new,
                 RankingExpressionTypeResolver::new,
                 // These should be last:
                 IndexingValidation::new,
diff --git a/config-model/src/main/java/com/yahoo/vespa/model/application/validation/RankSetupValidator.java b/config-model/src/main/java/com/yahoo/vespa/model/application/validation/RankSetupValidator.java
index c6c7969e466..d5c5183b01f 100644
--- a/config-model/src/main/java/com/yahoo/vespa/model/application/validation/RankSetupValidator.java
+++ b/config-model/src/main/java/com/yahoo/vespa/model/application/validation/RankSetupValidator.java
@@ -1,20 +1,19 @@
 // Copyright 2017 Yahoo Holdings. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
 package com.yahoo.vespa.model.application.validation;
 
-import com.yahoo.cloud.config.ConfigserverConfig;
 import com.yahoo.config.model.deploy.DeployState;
 import com.yahoo.io.IOUtils;
 import com.yahoo.log.InvalidLogFormatException;
 import java.util.logging.Level;
 import com.yahoo.log.LogMessage;
 import com.yahoo.searchdefinition.OnnxModel;
+import com.yahoo.searchdefinition.processing.OnnxModelTypeResolver;
 import com.yahoo.yolean.Exceptions;
 import com.yahoo.system.ProcessExecuter;
 import com.yahoo.text.StringUtilities;
 import com.yahoo.vespa.config.search.AttributesConfig;
 import com.yahoo.collections.Pair;
 import com.yahoo.config.ConfigInstance;
-import com.yahoo.vespa.defaults.Defaults;
 import com.yahoo.vespa.config.search.ImportedFieldsConfig;
 import com.yahoo.vespa.config.search.IndexschemaConfig;
 import com.yahoo.vespa.config.search.RankProfilesConfig;
@@ -31,7 +30,6 @@ import com.yahoo.vespa.model.search.SearchCluster;
 import java.io.File;
 import java.io.IOException;
 import java.nio.file.Files;
-import java.nio.file.Paths;
 import java.time.Duration;
 import java.time.Instant;
 import java.util.logging.Logger;
@@ -152,12 +150,9 @@ public class RankSetupValidator extends Validator {
         // Assist verify-ranksetup in finding the actual ONNX model files
         Map<String, OnnxModel> models = db.getDerivedConfiguration().getSearch().onnxModels().asMap();
         if (models.values().size() > 0) {
-            ConfigserverConfig cfg = new ConfigserverConfig(new ConfigserverConfig.Builder());  // assume defaults
-            String fileRefDir = Defaults.getDefaults().underVespaHome(cfg.fileReferencesDir());
             List<String> config = new ArrayList<>(models.values().size() * 2);
             for (OnnxModel model : models.values()) {
-                String modelFilename = Paths.get(model.getFileName()).getFileName().toString();
-                String modelPath = Paths.get(fileRefDir, model.getFileReference(), modelFilename).toString();
+                String modelPath = OnnxModelTypeResolver.getFileRepositoryPath(model.getFilePath(), model.getFileReference());
                 config.add(String.format("file[%d].ref \"%s\"", config.size() / 2, model.getFileReference()));
                 config.add(String.format("file[%d].path \"%s\"", config.size() / 2, modelPath));
             }
diff --git a/config-model/src/main/java/com/yahoo/vespa/model/ml/ConvertedModel.java b/config-model/src/main/java/com/yahoo/vespa/model/ml/ConvertedModel.java
index 943fcbf6c1d..5ee6ed02e61 100644
--- a/config-model/src/main/java/com/yahoo/vespa/model/ml/ConvertedModel.java
+++ b/config-model/src/main/java/com/yahoo/vespa/model/ml/ConvertedModel.java
@@ -150,7 +150,7 @@ public class ConvertedModel {
      */
     public ExpressionNode expression(FeatureArguments arguments, RankProfileTransformContext context) {
         ExpressionFunction expression = selectExpression(arguments);
-        if (sourceModel.isPresent()) // we should verify
+        if (sourceModel.isPresent() && context != null) // we should verify
             verifyInputs(expression.getBody(), sourceModel.get(), context.rankProfile(), context.queryProfiles());
         return expression.getBody().getRoot();
     }
diff --git a/config-model/src/main/protobuf/onnx.proto b/config-model/src/main/protobuf/onnx.proto
new file mode 100644
index 00000000000..dc6542867e0
--- /dev/null
+++ b/config-model/src/main/protobuf/onnx.proto
@@ -0,0 +1,464 @@
+//
+// WARNING: This file is automatically generated!  Please edit onnx.in.proto.
+//
+
+
+// Copyright (c) Facebook Inc. and Microsoft Corporation.
+// Licensed under the MIT license.
+
+syntax = "proto2";
+
+package onnx;
+
+// Overview
+//
+// ONNX is an open specification that is comprised of the following components:
+//
+// 1)  A definition of an extensible computation graph model.
+// 2)  Definitions of standard data types.
+// 3)  Definitions of built-in operators.
+//
+// This document describes the syntax of models and their computation graphs,
+// as well as the standard data types. Together, they are referred to as the ONNX
+// Intermediate Representation, or 'IR' for short. 
+//
+// The normative semantic specification of the ONNX IR is found in docs/IR.md.
+// Definitions of the built-in neural network operators may be found in docs/Operators.md.
+
+// Notes
+//
+// Release
+//
+// We are still in the very early stage of defining ONNX. The current
+// version of ONNX is a starting point. While we are actively working
+// towards a complete spec, we would like to get the community involved
+// by sharing our working version of ONNX.
+//
+// Protobuf compatibility
+// 
+// To simplify framework compatibility, ONNX is defined using the subset of protobuf 
+// that is compatible with both protobuf v2 and v3. This means that we do not use any
+// protobuf features that are only available in one of the two versions.
+//
+// Here are the most notable contortions we have to carry out to work around
+// these limitations:
+//
+//   - No 'map' (added protobuf 3.0). We instead represent mappings as lists
+//     of key-value pairs, where order does not matter and duplicates
+//     are not allowed.
+
+
+// Versioning
+//
+// ONNX versioning is specified in docs/IR.md and elaborated on in docs/Versioning.md
+//
+// To be compatible with both proto2 and proto3, we will use a version number
+// that is not defined by the default value but an explicit enum number.
+enum Version {
+  // proto3 requires the first enum value to be zero.
+  // We add this just to appease the compiler.
+  _START_VERSION = 0;
+  // The version field is always serialized and we will use it to store the
+  // version that the  graph is generated from. This helps us set up version
+  // control. We should use version as
+  //     xx(major) - xx(minor) - xxxx(bugfix)
+  // and we are starting with 0x00000001 (0.0.1), which was the
+  //  version we published on Oct 10, 2017.
+  IR_VERSION_2017_10_10 = 0x00000001;
+
+  // IR_VERSION 0.0.2 published on Oct 30, 2017
+  // - Added type discriminator to AttributeProto to support proto3 users
+  IR_VERSION_2017_10_30 = 0x00000002;
+
+  // IR VERSION 0.0.3 published on Nov 3, 2017
+  // - For operator versioning:
+  //    - Added new message OperatorSetIdProto
+  //    - Added opset_import in ModelProto
+  // - For vendor extensions, added domain in NodeProto
+  IR_VERSION = 0x00000003;
+}
+
+// Attributes
+//
+// A named attribute containing either singular float, integer, string, graph,
+// and tensor values, or repeated float, integer, string, graph, and tensor values.
+// An AttributeProto MUST contain the name field, and *only one* of the
+// following content fields, effectively enforcing a C/C++ union equivalent.
+message AttributeProto {
+
+  // Note: this enum is structurally identical to the OpSchema::AttrType
+  // enum defined in schema.h.  If you rev one, you likely need to rev the other.
+  enum AttributeType {
+    UNDEFINED = 0;
+    FLOAT = 1;
+    INT = 2;
+    STRING = 3;
+    TENSOR = 4;
+    GRAPH = 5;
+
+    FLOATS = 6;
+    INTS = 7;
+    STRINGS = 8;
+    TENSORS = 9;
+    GRAPHS = 10;
+  }
+
+  // The name field MUST be present for this version of the IR.
+  optional string name = 1;           // namespace Attribute
+ 
+  // if ref_attr_name is not empty, ref_attr_name is the attribute name in parent function.
+  // In this case, this AttributeProto does not contain data, and it's a reference of attribute
+  // in parent scope.
+  // NOTE: This should ONLY be used in function (sub-graph). It's invalid to be used in main graph.
+  optional string ref_attr_name = 21;
+
+  // A human-readable documentation for this attribute. Markdown is allowed.
+  optional string doc_string = 13;
+
+  // The type field MUST be present for this version of the IR.
+  // For 0.0.1 versions of the IR, this field was not defined, and
+  // implementations needed to use has_field hueristics to determine
+  // which value field was in use.  For IR_VERSION 0.0.2 or later, this
+  // field MUST be set and match the f|i|s|t|... field in use.  This
+  // change was made to accomodate proto3 implementations.
+  optional AttributeType type = 20;   // discriminator that indicates which field below is in use
+
+  // Exactly ONE of the following fields must be present for this version of the IR
+  optional float f = 2;               // float
+  optional int64 i = 3;               // int
+  optional bytes s = 4;               // UTF-8 string
+  optional TensorProto t = 5;         // tensor value
+  optional GraphProto g = 6;          // graph
+  // Do not use field below, it's deprecated.
+  // optional ValueProto v = 12;         // value - subsumes everything but graph
+
+  repeated float floats = 7;          // list of floats
+  repeated int64 ints = 8;            // list of ints
+  repeated bytes strings = 9;         // list of UTF-8 strings
+  repeated TensorProto tensors = 10;  // list of tensors
+  repeated GraphProto graphs = 11;    // list of graph
+}
+
+// Defines information on value, including the name, the type, and
+// the shape of the value.
+message ValueInfoProto {
+  // This field MUST be present in this version of the IR.
+  optional string name = 1;     // namespace Value
+  // This field MUST be present in this version of the IR.
+  optional TypeProto type = 2;
+  // A human-readable documentation for this value. Markdown is allowed.
+  optional string doc_string = 3;
+}
+
+// Nodes
+//
+// Computation graphs are made up of a DAG of nodes, which represent what is
+// commonly called a "layer" or "pipeline stage" in machine learning frameworks.
+//
+// For example, it can be a node of type "Conv" that takes in an image, a filter 
+// tensor and a bias tensor, and produces the convolved output.
+message NodeProto {
+  repeated string input = 1;    // namespace Value
+  repeated string output = 2;   // namespace Value
+
+  // An optional identifier for this node in a graph.
+  // This field MAY be absent in ths version of the IR.
+  optional string name = 3;     // namespace Node
+
+  // The symbolic identifier of the Operator to execute.
+  optional string op_type = 4;  // namespace Operator
+  // The domain of the OperatorSet that specifies the operator named by op_type.
+  optional string domain = 7;   // namespace Domain
+
+  // Additional named attributes.
+  repeated AttributeProto attribute = 5;
+
+  // A human-readable documentation for this node. Markdown is allowed.
+  optional string doc_string = 6;
+}
+
+// Models
+//
+// ModelProto is a top-level file/container format for bundling a ML model and
+// associating its computation graph with metadata.
+//
+// The semantics of the model are described by the associated GraphProto.
+message ModelProto {
+  // The version of the IR this model targets. See Version enum above.
+  // This field MUST be present.
+  optional int64 ir_version = 1;
+
+  // The OperatorSets this model relies on.
+  // All ModelProtos MUST have at least one entry that
+  // specifies which version of the ONNX OperatorSet is
+  // being imported.
+  //
+  // All nodes in the ModelProto's graph will bind against the operator
+  // with the same-domain/same-op_type operator with the HIGHEST version
+  // in the referenced operator sets.
+  repeated OperatorSetIdProto opset_import = 8;
+
+  // The name of the framework or tool used to generate this model.
+  // This field SHOULD be present to indicate which implementation/tool/framework
+  // emitted the model.
+  optional string producer_name = 2;
+
+  // The version of the framework or tool used to generate this model.
+  // This field SHOULD be present to indicate which implementation/tool/framework
+  // emitted the model.
+  optional string producer_version = 3;
+
+  // Domain name of the model.
+  // We use reverse domain names as name space indicators. For example:
+  // `com.facebook.fair` or `com.microsoft.cognitiveservices`
+  //
+  // Together with `model_version` and GraphProto.name, this forms the unique identity of
+  // the graph.
+  optional string domain = 4;
+
+  // The version of the graph encoded. See Version enum below.
+  optional int64 model_version = 5;
+
+  // A human-readable documentation for this model. Markdown is allowed.
+  optional string doc_string = 6;
+
+  // The parameterized graph that is evaluated to execute the model.
+  optional GraphProto graph = 7;
+
+  // Named metadata values; keys should be distinct.
+  repeated StringStringEntryProto metadata_props = 14;
+};
+
+// StringStringEntryProto follows the pattern for cross-proto-version maps.
+// See https://developers.google.com/protocol-buffers/docs/proto3#maps
+message StringStringEntryProto {
+  optional string key = 1;
+  optional string value= 2;
+};
+
+// Graphs
+//
+// A graph defines the computational logic of a model and is comprised of a parameterized 
+// list of nodes that form a directed acyclic graph based on their inputs and outputs.
+// This is the equivalent of the "network" or "graph" in many deep learning
+// frameworks.
+message GraphProto {
+  // The nodes in the graph, sorted topologically.
+  repeated NodeProto node = 1;
+
+  // The name of the graph.
+  optional string name = 2;   // namespace Graph
+
+  // A list of named tensor values, used to specify constant inputs of the graph.
+  // Each TensorProto entry must have a distinct name (within the list) that
+  // also appears in the input list.
+  repeated TensorProto initializer = 5;
+
+  // A human-readable documentation for this graph. Markdown is allowed.
+  optional string doc_string = 10;
+
+  // The inputs and outputs of the graph.
+  repeated ValueInfoProto input = 11;
+  repeated ValueInfoProto output = 12;
+
+  // Information for the values in the graph. The ValueInfoProto.name's
+  // must be distinct. It is optional for a value to appear in value_info list.
+  repeated ValueInfoProto value_info = 13;
+
+  // DO NOT USE the following fields, they were deprecated from earlier versions.
+  // repeated string input = 3;
+  // repeated string output = 4;
+  // optional int64 ir_version = 6;
+  // optional int64 producer_version = 7;
+  // optional string producer_tag = 8;
+  // optional string domain = 9;
+}
+
+// Tensors
+//
+// A serialized tensor value.
+message TensorProto {
+  enum DataType {
+    UNDEFINED = 0;
+    // Basic types.
+    FLOAT = 1;   // float
+    UINT8 = 2;   // uint8_t
+    INT8 = 3;    // int8_t
+    UINT16 = 4;  // uint16_t
+    INT16 = 5;   // int16_t
+    INT32 = 6;   // int32_t
+    INT64 = 7;   // int64_t
+    STRING = 8;  // string
+    BOOL = 9;    // bool
+
+    // Advanced types
+    FLOAT16 = 10;
+    DOUBLE = 11;
+    UINT32 = 12;
+    UINT64 = 13;
+    COMPLEX64 = 14;     // complex with float32 real and imaginary components
+    COMPLEX128 = 15;    // complex with float64 real and imaginary components
+    // Future extensions go here.
+  }
+
+  // The shape of the tensor.
+  repeated int64 dims = 1;
+
+  // The data type of the tensor.
+  optional DataType data_type = 2;
+
+  // For very large tensors, we may want to store them in chunks, in which
+  // case the following fields will specify the segment that is stored in
+  // the current TensorProto.
+  message Segment {
+    optional int64 begin = 1;
+    optional int64 end = 2;
+  }
+  optional Segment segment = 3;
+
+  // Tensor content must be organized in row-major order.
+  //
+  // Depending on the data_type field, exactly one of the fields below with
+  // name ending in _data is used to store the elements of the tensor.
+
+  // For float and complex64 values
+  // Complex64 tensors are encoded as a single array of floats,
+  // with the real components appearing in odd numbered positions,
+  // and the corresponding imaginary component apparing in the
+  // subsequent even numbered position. (e.g., [1.0 + 2.0i, 3.0 + 4.0i]
+  // is encoded as [1.0, 2.0 ,3.0 ,4.0]
+  // When this field is present, the data_type field MUST be FLOAT or COMPLEX64.
+  repeated float float_data = 4 [packed = true];
+
+  // For int32, uint8, int8, uint16, int16, bool, and float16 values
+  // float16 values must be bit-wise converted to an uint16_t prior
+  // to writing to the buffer.
+  // When this field is present, the data_type field MUST be
+  // INT32, INT16, INT8, UINT16, INT8, BOOL, or FLOAT32
+  repeated int32 int32_data = 5 [packed = true];
+
+  // For strings.
+  // Each element of string_data is a UTF-8 encoded Unicode
+  // string. No trailing null, no leading BOM. The protobuf "string"
+  // scalar type is not used to match ML community conventions.
+  // When this field is present, the data_type field MUST be STRING
+  repeated bytes string_data = 6;
+
+  // For int64.
+  // When this field is present, the data_type field MUST be INT64
+  repeated int64 int64_data = 7 [packed = true];
+
+  // Optionally, a name for the tensor.
+  optional string name = 8; // namespace Value
+
+  // A human-readable documentation for this tensor. Markdown is allowed.
+  optional string doc_string = 12;
+
+  // Serializations can either use one of the fields above, or use this
+  // raw bytes field. The only exception is the string case, where one is
+  // required to store the content in the repeated bytes string_data field.
+  //
+  // When this raw_data field is used to store tensor value, elements MUST
+  // be stored in as fixed-width, little-endian order.
+  // Floating-point data types MUST be stored in IEEE 754 format.
+  // Complex64 elements must be written as two consecutive FLOAT values, real component first.
+  // Complex128 elements must be written as two consecutive DOUBLE values, real component first.
+  // Boolean type MUST be written one byte per tensor element (00000001 for true, 00000000 for false).
+  //
+  // Note: the advantage of specific field rather than the raw_data field is
+  // that in some cases (e.g. int data), protobuf does a better packing via
+  // variable length storage, and may lead to smaller binary footprint.
+  // When this field is present, the data_type field MUST NOT be STRING or UNDEFINED
+  optional bytes raw_data = 9;
+
+  // For double
+  // Complex64 tensors are encoded as a single array of doubles,
+  // with the real components appearing in odd numbered positions,
+  // and the corresponding imaginary component apparing in the
+  // subsequent even numbered position. (e.g., [1.0 + 2.0i, 3.0 + 4.0i]
+  // is encoded as [1.0, 2.0 ,3.0 ,4.0]
+  // When this field is present, the data_type field MUST be DOUBLE or COMPLEX128
+  repeated double double_data = 10 [packed = true];
+
+  // For uint64 and uint32 values
+  // When this field is present, the data_type field MUST be
+  // UINT32 or UINT64
+  repeated uint64 uint64_data = 11 [packed = true];
+}
+
+// Defines a tensor shape. A dimension can be either an integer value
+// or a symbolic variable. A symbolic variable represents an unknown
+// dimension.
+message TensorShapeProto {
+  message Dimension {
+    oneof value {
+      int64 dim_value = 1;
+      string dim_param = 2;   // namespace Shape
+    };
+    // Standard denotation can optionally be used to denote tensor
+    // dimensions with standard semantic descriptions to ensure
+    // that operations are applied to the correct axis of a tensor.
+    optional string denotation = 3;
+  };
+  repeated Dimension dim = 1;
+}
+
+// A set of pre-defined constants to be used as values for
+// the standard denotation field in TensorShapeProto.Dimension
+// for semantic description of the tensor dimension.
+message DenotationConstProto {
+  // Describe a batch number dimension.
+  optional string DATA_BATCH = 1 [default = "DATA_BATCH"];
+  // Describe a channel dimension.
+  optional string DATA_CHANNEL = 2 [default = "DATA_CHANNEL"];
+  // Describe a time dimension.
+  optional string DATA_TIME = 3 [default = "DATA_TIME"];
+  // Describe a feature dimension. This is typically a feature
+  // dimension in RNN and/or spatial dimension in CNN.
+  optional string DATA_FEATURE = 4 [default = "DATA_FEATURE"];
+  // Describe a filter in-channel dimension. This is the dimension
+  // that is identical (in size) to the channel dimension of the input
+  // image feature maps.
+  optional string FILTER_IN_CHANNEL = 5 [default = "FILTER_IN_CHANNEL"];
+  // Describe a filter out channel dimension. This is the dimension
+  // that is identical (int size) to the channel dimension of the output
+  // image feature maps.
+  optional string FILTER_OUT_CHANNEL = 6 [default = "FILTER_OUT_CHANNEL"];
+  // Describe a filter spatial dimension.
+  optional string FILTER_SPATIAL = 7 [default = "FILTER_SPATIAL"];
+}
+
+// Types
+//
+// The standard ONNX data types.
+message TypeProto {
+
+  message Tensor {
+    // This field MUST NOT have the value of UNDEFINED
+    // This field MUST be present for this version of the IR.
+    optional TensorProto.DataType elem_type = 1;
+    optional TensorShapeProto shape = 2;
+  }
+
+
+  oneof value {
+    // The type of a tensor.
+    Tensor tensor_type = 1;
+
+  }
+}
+
+// Operator Sets
+//
+// OperatorSets are uniquely identified by a (domain, opset_version) pair.
+message OperatorSetIdProto {
+  // The domain of the operator set being identified.
+  // The empty string ("") or absence of this field implies the operator
+  // set that is defined as part of the ONNX specification.
+  // This field MUST be present in this version of the IR when referring to any other operator set.
+  optional string domain = 1;
+
+  // The version of the operator set being identified.
+  // This field MUST be present in this version of the IR.
+  optional int64 version = 2;
+}
\ No newline at end of file
diff --git a/config-model/src/test/cfg/application/ml_models/searchdefinitions/test.sd b/config-model/src/test/cfg/application/ml_models/searchdefinitions/test.sd
index e9575af6010..cc73f2daff5 100644
--- a/config-model/src/test/cfg/application/ml_models/searchdefinitions/test.sd
+++ b/config-model/src/test/cfg/application/ml_models/searchdefinitions/test.sd
@@ -18,7 +18,7 @@ search test {
         }
 
         function mnist_softmax_onnx() {
-            expression: onnx("mnist_softmax")
+            expression: onnx_vespa("mnist_softmax")
         }
 
         function my_xgboost() {
diff --git a/config-model/src/test/integration/onnx-model/files/create_dynamic_model.py b/config-model/src/test/integration/onnx-model/files/create_dynamic_model.py
new file mode 100755
index 00000000000..55df3a557e9
--- /dev/null
+++ b/config-model/src/test/integration/onnx-model/files/create_dynamic_model.py
@@ -0,0 +1,12 @@
+# Copyright Verizon Media. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+
+import onnx
+from onnx import helper, TensorProto
+
+INPUT = helper.make_tensor_value_info('input', TensorProto.FLOAT, ["batch", "sequence"])
+OUTPUT = helper.make_tensor_value_info('output', TensorProto.FLOAT, ["batch", "sequence"])
+
+nodes = [helper.make_node('Identity', ['input'], ['output'])]
+graph_def = helper.make_graph( nodes, 'simple_scoring', [INPUT], [OUTPUT])
+model_def = helper.make_model(graph_def, producer_name='create_dynamic_model.py')
+onnx.save(model_def, 'dynamic_model.onnx')
diff --git a/config-model/src/test/integration/onnx-model/files/create_model.py b/config-model/src/test/integration/onnx-model/files/create_model.py
new file mode 100755
index 00000000000..10ff92c2eda
--- /dev/null
+++ b/config-model/src/test/integration/onnx-model/files/create_model.py
@@ -0,0 +1,37 @@
+# Copyright Verizon Media. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+
+import onnx
+from onnx import helper, TensorProto
+
+INPUT_1 = helper.make_tensor_value_info('first_input', TensorProto.FLOAT, [2])
+INPUT_2 = helper.make_tensor_value_info('second/input:0', TensorProto.FLOAT, [2])
+INPUT_3 = helper.make_tensor_value_info('third_input', TensorProto.FLOAT, [2])
+OUTPUT_1 = helper.make_tensor_value_info('path/to/output:0', TensorProto.FLOAT, [2])
+OUTPUT_2 = helper.make_tensor_value_info('path/to/output:1', TensorProto.FLOAT, [2])
+OUTPUT_3 = helper.make_tensor_value_info('path/to/output:2', TensorProto.FLOAT, [2])
+
+nodes = [
+    helper.make_node(
+        'Add',
+        ['first_input', 'second/input:0'],
+        ['path/to/output:0'],
+    ),
+    helper.make_node(
+        'Add',
+        ['third_input', 'second/input:0'],
+        ['path/to/output:1']
+    ),
+    helper.make_node(
+        'Add',
+        ['path/to/output:0', 'path/to/output:1'],
+        ['path/to/output:2']
+    ),
+]
+graph_def = helper.make_graph(
+    nodes,
+    'simple_scoring',
+    [INPUT_1, INPUT_2, INPUT_3],
+    [OUTPUT_1, OUTPUT_2, OUTPUT_3]
+)
+model_def = helper.make_model(graph_def, producer_name='create_model.py')
+onnx.save(model_def, 'model.onnx')
diff --git a/config-model/src/test/integration/onnx-model/files/dynamic_model.onnx b/config-model/src/test/integration/onnx-model/files/dynamic_model.onnx
new file mode 100644
index 00000000000..6bbdad2d76e
--- /dev/null
+++ b/config-model/src/test/integration/onnx-model/files/dynamic_model.onnx
@@ -0,0 +1,13 @@
+create_dynamic_model.py:x
+
+inputoutput"Identitysimple_scoringZ$
+input
+

+batch
+
+sequenceb%
+output
+

+batch
+
+sequenceB
\ No newline at end of file
diff --git a/config-model/src/test/integration/onnx-model/files/model.onnx b/config-model/src/test/integration/onnx-model/files/model.onnx
new file mode 100644
index 00000000000..f3898205c6a
--- /dev/null
+++ b/config-model/src/test/integration/onnx-model/files/model.onnx
@@ -0,0 +1,34 @@
+create_model.py:í
+4
+first_input
+second/input:0path/to/output:0"Add
+4
+third_input
+second/input:0path/to/output:1"Add
+;
+path/to/output:0
+path/to/output:1path/to/output:2"Addsimple_scoringZ
+first_input
+
+

+Z
+second/input:0
+
+

+Z
+third_input
+
+

+b
+path/to/output:0
+
+

+b
+path/to/output:1
+
+

+b
+path/to/output:2
+
+

+B
\ No newline at end of file
diff --git a/config-model/src/test/integration/onnx-model/files/summary_model.onnx b/config-model/src/test/integration/onnx-model/files/summary_model.onnx
new file mode 100644
index 00000000000..f3898205c6a
--- /dev/null
+++ b/config-model/src/test/integration/onnx-model/files/summary_model.onnx
@@ -0,0 +1,34 @@
+create_model.py:í
+4
+first_input
+second/input:0path/to/output:0"Add
+4
+third_input
+second/input:0path/to/output:1"Add
+;
+path/to/output:0
+path/to/output:1path/to/output:2"Addsimple_scoringZ
+first_input
+
+

+Z
+second/input:0
+
+

+Z
+third_input
+
+

+b
+path/to/output:0
+
+

+b
+path/to/output:1
+
+

+b
+path/to/output:2
+
+

+B
\ No newline at end of file
diff --git a/config-model/src/test/integration/onnx-model/searchdefinitions/test.sd b/config-model/src/test/integration/onnx-model/searchdefinitions/test.sd
index 0f0fa694e6f..6e9ba356293 100644
--- a/config-model/src/test/integration/onnx-model/searchdefinitions/test.sd
+++ b/config-model/src/test/integration/onnx-model/searchdefinitions/test.sd
@@ -14,7 +14,7 @@ search test {
     }
 
     onnx-model my_model {
-        file:  files/ranking_model.onnx
+        file:  files/model.onnx
         input  first_input: attribute(document_field)
         input  "second/input:0": constant(my_constant)
         input  "third_input": my_function
@@ -22,19 +22,25 @@ search test {
     }
 
     onnx-model another_model {
-        file:  files/ranking_model.onnx
+        file:  files/model.onnx
         input  first_input: attribute(document_field)
         input  "second/input:0": constant(my_constant)
         input  "third_input": another_function
         output "path/to/output:2": out
     }
 
+    onnx-model dynamic_model {
+        file:  files/dynamic_model.onnx
+        input  input: my_function
+        output output: my_output
+    }
+
     rank-profile test_model_config {
         function my_function() {
             expression: tensor(d0[2])(1)
         }
         first-phase {
-            expression: onnxModel(my_model).out
+            expression: onnxModel(my_model).out{d0:1}
         }
     }
 
@@ -49,7 +55,7 @@ search test {
             expression: my_function()
         }
         first-phase {
-            expression: onnxModel("files/ranking_model.onnx", "path/to/output:1")
+            expression: onnxModel("files/model.onnx", "path/to/output:1"){d0:1}
         }
     }
 
@@ -62,9 +68,29 @@ search test {
         }
         summary-features {
             onnxModel(another_model).out
-            onnxModel("files/ranking_model.onnx", "path/to/output:2")
+            onnxModel("files/summary_model.onnx", "path/to/output:2")
+        }
+    }
+
+    rank-profile test_dynamic_model {
+        function my_function() {
+            expression: tensor(d0[1],d1[2])(d1)
+        }
+        first-phase {
+            expression: onnxModel(dynamic_model){d0:0,d1:1}
         }
+    }
 
+    rank-profile test_dynamic_model_2 {
+        function my_function_2() {
+            expression: tensor(d0[1],d1[3])(d1)
+        }
+        function my_function() {
+            expression: my_function_2()
+        }
+        first-phase {
+            expression: onnxModel(dynamic_model){d0:0,d1:2}
+        }
     }
 
 }
diff --git a/config-model/src/test/java/com/yahoo/searchdefinition/processing/RankingExpressionWithOnnxModelTestCase.java b/config-model/src/test/java/com/yahoo/searchdefinition/processing/RankingExpressionWithOnnxModelTestCase.java
index d9b0c70dfdd..5060aafb55f 100644
--- a/config-model/src/test/java/com/yahoo/searchdefinition/processing/RankingExpressionWithOnnxModelTestCase.java
+++ b/config-model/src/test/java/com/yahoo/searchdefinition/processing/RankingExpressionWithOnnxModelTestCase.java
@@ -25,43 +25,61 @@ public class RankingExpressionWithOnnxModelTestCase {
         OnnxModelsConfig.Builder builder = new OnnxModelsConfig.Builder();
         ((OnnxModelsConfig.Producer) db).getConfig(builder);
         OnnxModelsConfig config = new OnnxModelsConfig(builder);
-        assertEquals(3, config.model().size());
+        assertEquals(5, config.model().size());
 
-        assertEquals("my_model", config.model(1).name());
+        assertEquals("my_model", config.model(0).name());
+        assertEquals(3, config.model(0).input().size());
+        assertEquals("second/input:0", config.model(0).input(0).name());
+        assertEquals("constant(my_constant)", config.model(0).input(0).source());
+        assertEquals("first_input", config.model(0).input(1).name());
+        assertEquals("attribute(document_field)", config.model(0).input(1).source());
+        assertEquals("third_input", config.model(0).input(2).name());
+        assertEquals("rankingExpression(my_function)", config.model(0).input(2).source());
+        assertEquals(3, config.model(0).output().size());
+        assertEquals("path/to/output:0", config.model(0).output(0).name());
+        assertEquals("out", config.model(0).output(0).as());
+        assertEquals("path/to/output:1", config.model(0).output(1).name());
+        assertEquals("path_to_output_1", config.model(0).output(1).as());
+        assertEquals("path/to/output:2", config.model(0).output(2).name());
+        assertEquals("path_to_output_2", config.model(0).output(2).as());
+
+        assertEquals("files_model_onnx", config.model(1).name());
         assertEquals(3, config.model(1).input().size());
-        assertEquals("first_input", config.model(1).input(0).name());
-        assertEquals("attribute(document_field)", config.model(1).input(0).source());
-        assertEquals("second/input:0", config.model(1).input(1).name());
-        assertEquals("constant(my_constant)", config.model(1).input(1).source());
-        assertEquals("third_input", config.model(1).input(2).name());
-        assertEquals("rankingExpression(my_function)", config.model(1).input(2).source());
-        assertEquals(1, config.model(1).output().size());
+        assertEquals(3, config.model(1).output().size());
         assertEquals("path/to/output:0", config.model(1).output(0).name());
-        assertEquals("out", config.model(1).output(0).as());
-
-        assertEquals("files_ranking_model_onnx", config.model(0).name());
-        assertEquals(0, config.model(0).input().size());
-        assertEquals(2, config.model(0).output().size());
-        assertEquals("path/to/output:1", config.model(0).output(0).name());
-        assertEquals("path_to_output_1", config.model(0).output(0).as());
-        assertEquals("path/to/output:2", config.model(0).output(1).name());
-        assertEquals("path_to_output_2", config.model(0).output(1).as());
+        assertEquals("path_to_output_0", config.model(1).output(0).as());
+        assertEquals("path/to/output:1", config.model(1).output(1).name());
+        assertEquals("path_to_output_1", config.model(1).output(1).as());
+        assertEquals("path/to/output:2", config.model(1).output(2).name());
+        assertEquals("path_to_output_2", config.model(1).output(2).as());
+        assertEquals("files_model_onnx", config.model(1).name());
 
         assertEquals("another_model", config.model(2).name());
         assertEquals("third_input", config.model(2).input(2).name());
         assertEquals("rankingExpression(another_function)", config.model(2).input(2).source());
+
+        assertEquals("files_summary_model_onnx", config.model(3).name());
+        assertEquals(3, config.model(3).input().size());
+        assertEquals(3, config.model(3).output().size());
+
+        assertEquals("dynamic_model", config.model(4).name());
+        assertEquals(1, config.model(4).input().size());
+        assertEquals(1, config.model(4).output().size());
+        assertEquals("rankingExpression(my_function)", config.model(4).input(0).source());
     }
 
     private void assertTransformedFeature(DocumentDatabase db) {
         RankProfilesConfig.Builder builder = new RankProfilesConfig.Builder();
         ((RankProfilesConfig.Producer) db).getConfig(builder);
         RankProfilesConfig config = new RankProfilesConfig(builder);
-        assertEquals(5, config.rankprofile().size());
+        assertEquals(7, config.rankprofile().size());
 
         assertEquals("test_model_config", config.rankprofile(2).name());
         assertEquals("rankingExpression(my_function).rankingScript", config.rankprofile(2).fef().property(0).name());
         assertEquals("vespa.rank.firstphase", config.rankprofile(2).fef().property(2).name());
-        assertEquals("onnxModel(my_model).out", config.rankprofile(2).fef().property(2).value());
+        assertEquals("rankingExpression(firstphase)", config.rankprofile(2).fef().property(2).value());
+        assertEquals("rankingExpression(firstphase).rankingScript", config.rankprofile(2).fef().property(3).name());
+        assertEquals("onnxModel(my_model).out{d0:1}", config.rankprofile(2).fef().property(3).value());
 
         assertEquals("test_generated_model_config", config.rankprofile(3).name());
         assertEquals("rankingExpression(my_function).rankingScript", config.rankprofile(3).fef().property(0).name());
@@ -69,16 +87,28 @@ public class RankingExpressionWithOnnxModelTestCase {
         assertEquals("rankingExpression(second_input).rankingScript", config.rankprofile(3).fef().property(4).name());
         assertEquals("rankingExpression(third_input).rankingScript", config.rankprofile(3).fef().property(6).name());
         assertEquals("vespa.rank.firstphase", config.rankprofile(3).fef().property(8).name());
-        assertEquals("onnxModel(files_ranking_model_onnx).path_to_output_1", config.rankprofile(3).fef().property(8).value());
+        assertEquals("rankingExpression(firstphase)", config.rankprofile(3).fef().property(8).value());
+        assertEquals("rankingExpression(firstphase).rankingScript", config.rankprofile(3).fef().property(9).name());
+        assertEquals("onnxModel(files_model_onnx).path_to_output_1{d0:1}", config.rankprofile(3).fef().property(9).value());
 
         assertEquals("test_summary_features", config.rankprofile(4).name());
         assertEquals("rankingExpression(another_function).rankingScript", config.rankprofile(4).fef().property(0).name());
         assertEquals("rankingExpression(firstphase).rankingScript", config.rankprofile(4).fef().property(3).name());
         assertEquals("1", config.rankprofile(4).fef().property(3).value());
         assertEquals("vespa.summary.feature", config.rankprofile(4).fef().property(4).name());
-        assertEquals("onnxModel(another_model).out", config.rankprofile(4).fef().property(4).value());
+        assertEquals("onnxModel(files_summary_model_onnx).path_to_output_2", config.rankprofile(4).fef().property(4).value());
         assertEquals("vespa.summary.feature", config.rankprofile(4).fef().property(5).name());
-        assertEquals("onnxModel(files_ranking_model_onnx).path_to_output_2", config.rankprofile(4).fef().property(5).value());
+        assertEquals("onnxModel(another_model).out", config.rankprofile(4).fef().property(5).value());
+
+        assertEquals("test_dynamic_model", config.rankprofile(5).name());
+        assertEquals("rankingExpression(my_function).rankingScript", config.rankprofile(5).fef().property(0).name());
+        assertEquals("rankingExpression(firstphase).rankingScript", config.rankprofile(5).fef().property(3).name());
+        assertEquals("onnxModel(dynamic_model).my_output{d0:0, d1:1}", config.rankprofile(5).fef().property(3).value());
+
+        assertEquals("test_dynamic_model_2", config.rankprofile(6).name());
+        assertEquals("rankingExpression(firstphase).rankingScript", config.rankprofile(6).fef().property(5).name());
+        assertEquals("onnxModel(dynamic_model).my_output{d0:0, d1:2}", config.rankprofile(6).fef().property(5).value());
+
     }
 
 }
diff --git a/config-model/src/test/java/com/yahoo/searchdefinition/processing/RankingExpressionWithOnnxTestCase.java b/config-model/src/test/java/com/yahoo/searchdefinition/processing/RankingExpressionWithOnnxTestCase.java
index 6bf69907609..40bf970a313 100644
--- a/config-model/src/test/java/com/yahoo/searchdefinition/processing/RankingExpressionWithOnnxTestCase.java
+++ b/config-model/src/test/java/com/yahoo/searchdefinition/processing/RankingExpressionWithOnnxTestCase.java
@@ -70,7 +70,7 @@ public class RankingExpressionWithOnnxTestCase {
     @Test
     public void testOnnxReferenceWithConstantFeature() {
         RankProfileSearchFixture search = fixtureWith("constant(mytensor)",
-                "onnx('mnist_softmax.onnx')",
+                "onnx_vespa('mnist_softmax.onnx')",
                 "constant mytensor { file: ignored\ntype: tensor<float>(d0[1],d1[784]) }",
                 null);
         search.assertFirstPhaseExpression(vespaExpression, "my_profile");
@@ -87,7 +87,7 @@ public class RankingExpressionWithOnnxTestCase {
                 queryProfile,
                 queryProfileType);
         RankProfileSearchFixture search = fixtureWith("query(mytensor)",
-                "onnx('mnist_softmax.onnx')",
+                "onnx_vespa('mnist_softmax.onnx')",
                 null,
                 null,
                 "Placeholder",
@@ -99,7 +99,7 @@ public class RankingExpressionWithOnnxTestCase {
     public void testOnnxReferenceWithDocumentFeature() {
         StoringApplicationPackage application = new StoringApplicationPackage(applicationDir);
         RankProfileSearchFixture search = fixtureWith("attribute(mytensor)",
-                "onnx('mnist_softmax.onnx')",
+                "onnx_vespa('mnist_softmax.onnx')",
                 null,
                 "field mytensor type tensor<float>(d0[1],d1[784]) { indexing: attribute }",
                 "Placeholder",
@@ -117,7 +117,7 @@ public class RankingExpressionWithOnnxTestCase {
                 "</query-profile-type>";
         StoringApplicationPackage application = new StoringApplicationPackage(applicationDir, queryProfile, queryProfileType);
         RankProfileSearchFixture search = fixtureWith("sum(query(mytensor) * attribute(mytensor) * constant(mytensor),d2)",
-                "onnx('mnist_softmax.onnx')",
+                "onnx_vespa('mnist_softmax.onnx')",
                 "constant mytensor { file: ignored\ntype: tensor<float>(d0[1],d1[784]) }",
                 "field mytensor type tensor<float>(d0[1],d1[784]) { indexing: attribute }",
                 "Placeholder",
@@ -129,21 +129,21 @@ public class RankingExpressionWithOnnxTestCase {
     @Test
     public void testNestedOnnxReference() {
         RankProfileSearchFixture search = fixtureWith("tensor<float>(d0[1],d1[784])(0.0)",
-                "5 + sum(onnx('mnist_softmax.onnx'))");
+                "5 + sum(onnx_vespa('mnist_softmax.onnx'))");
         search.assertFirstPhaseExpression("5 + reduce(" + vespaExpression + ", sum)", "my_profile");
     }
 
     @Test
     public void testOnnxReferenceWithSpecifiedOutput() {
         RankProfileSearchFixture search = fixtureWith("tensor<float>(d0[1],d1[784])(0.0)",
-                "onnx('mnist_softmax.onnx', 'layer_add')");
+                "onnx_vespa('mnist_softmax.onnx', 'layer_add')");
         search.assertFirstPhaseExpression(vespaExpression, "my_profile");
     }
 
     @Test
     public void testOnnxReferenceWithSpecifiedOutputAndSignature() {
         RankProfileSearchFixture search = fixtureWith("tensor<float>(d0[1],d1[784])(0.0)",
-                "onnx('mnist_softmax.onnx', 'default.layer_add')");
+                "onnx_vespa('mnist_softmax.onnx', 'default.layer_add')");
         search.assertFirstPhaseExpression(vespaExpression, "my_profile");
     }
 
@@ -155,7 +155,7 @@ public class RankingExpressionWithOnnxTestCase {
                     new QueryProfileRegistry(),
                     "  rank-profile my_profile {\n" +
                             "    first-phase {\n" +
-                            "      expression: onnx('mnist_softmax.onnx')" +
+                            "      expression: onnx_vespa('mnist_softmax.onnx')" +
                             "    }\n" +
                             "  }");
             search.compileRankProfile("my_profile", applicationDir.append("models"));
@@ -164,7 +164,7 @@ public class RankingExpressionWithOnnxTestCase {
         }
         catch (IllegalArgumentException expected) {
             assertEquals("Rank profile 'my_profile' is invalid: Could not use Onnx model from " +
-                            "onnx('mnist_softmax.onnx'): " +
+                            "onnx_vespa('mnist_softmax.onnx'): " +
                             "Model refers input 'Placeholder' of type tensor<float>(d0[1],d1[784]) but this function is " +
                             "not present in rank profile 'my_profile'",
                     Exceptions.toMessageString(expected));
@@ -175,13 +175,13 @@ public class RankingExpressionWithOnnxTestCase {
     public void testOnnxReferenceWithWrongFunctionType() {
         try {
             RankProfileSearchFixture search = fixtureWith("tensor(d0[1],d5[10])(0.0)",
-                    "onnx('mnist_softmax.onnx')");
+                    "onnx_vespa('mnist_softmax.onnx')");
             search.assertFirstPhaseExpression(vespaExpression, "my_profile");
             fail("Expecting exception");
         }
         catch (IllegalArgumentException expected) {
             assertEquals("Rank profile 'my_profile' is invalid: Could not use Onnx model from " +
-                            "onnx('mnist_softmax.onnx'): " +
+                            "onnx_vespa('mnist_softmax.onnx'): " +
                             "Model refers input 'Placeholder'. The required type of this is tensor<float>(d0[1],d1[784]), " +
                             "but this function returns tensor(d0[1],d5[10])",
                     Exceptions.toMessageString(expected));
@@ -192,13 +192,13 @@ public class RankingExpressionWithOnnxTestCase {
     public void testOnnxReferenceSpecifyingNonExistingOutput() {
         try {
             RankProfileSearchFixture search = fixtureWith("tensor<float>(d0[2],d1[784])(0.0)",
-                    "onnx('mnist_softmax.onnx', 'y')");
+                    "onnx_vespa('mnist_softmax.onnx', 'y')");
             search.assertFirstPhaseExpression(vespaExpression, "my_profile");
             fail("Expecting exception");
         }
         catch (IllegalArgumentException expected) {
             assertEquals("Rank profile 'my_profile' is invalid: Could not use Onnx model from " +
-                         "onnx('mnist_softmax.onnx','y'): " +
+                         "onnx_vespa('mnist_softmax.onnx','y'): " +
                          "No expressions named 'y' in model 'mnist_softmax.onnx'. Available expressions: default.layer_add",
                          Exceptions.toMessageString(expected));
         }
@@ -207,7 +207,7 @@ public class RankingExpressionWithOnnxTestCase {
     @Test
     public void testImportingFromStoredExpressions() throws IOException {
         RankProfileSearchFixture search = fixtureWith("tensor<float>(d0[1],d1[784])(0.0)",
-                "onnx('mnist_softmax.onnx')");
+                "onnx_vespa('mnist_softmax.onnx')");
         search.assertFirstPhaseExpression(vespaExpression, "my_profile");
 
         // At this point the expression is stored - copy application to another location which do not have a models dir
@@ -218,7 +218,7 @@ public class RankingExpressionWithOnnxTestCase {
                     storedApplicationDirectory.append(ApplicationPackage.MODELS_GENERATED_DIR).toFile());
             StoringApplicationPackage storedApplication = new StoringApplicationPackage(storedApplicationDirectory);
             RankProfileSearchFixture searchFromStored = fixtureWith("tensor<float>(d0[2],d1[784])(0.0)",
-                    "onnx('mnist_softmax.onnx')",
+                    "onnx_vespa('mnist_softmax.onnx')",
                     null,
                     null,
                     "Placeholder",
@@ -243,7 +243,7 @@ public class RankingExpressionWithOnnxTestCase {
                         "      expression: tensor<float>(d1[10],d2[784])(0.0)\n" +
                         "    }\n" +
                         "    first-phase {\n" +
-                        "      expression: onnx('mnist_softmax.onnx')" +
+                        "      expression: onnx_vespa('mnist_softmax.onnx')" +
                         "    }\n" +
                         "  }" +
                         "  rank-profile my_profile_child inherits my_profile {\n" +
@@ -288,7 +288,7 @@ public class RankingExpressionWithOnnxTestCase {
                         "      expression: tensor<float>(d0[3])(0.0)\n" +
                         "    }\n" +
                         "    first-phase {\n" +
-                        "      expression: onnx('" + name + ".onnx')" +
+                        "      expression: onnx_vespa('" + name + ".onnx')" +
                         "    }\n" +
                         "  }";
         final String functionName = "imported_ml_function_" + name + "_exp_output";
@@ -310,7 +310,7 @@ public class RankingExpressionWithOnnxTestCase {
                         "      expression: tensor<float>(d0[3])(0.0)\n" +
                         "    }\n" +
                         "    first-phase {\n" +
-                        "      expression: onnx('" + name + ".onnx')" +
+                        "      expression: onnx_vespa('" + name + ".onnx')" +
                         "    }\n" +
                         "  }" +
                         "  rank-profile my_profile_child inherits my_profile {\n" +
diff --git a/fat-model-dependencies/pom.xml b/fat-model-dependencies/pom.xml
index 4beaf6086a6..181ef6dffbd 100644
--- a/fat-model-dependencies/pom.xml
+++ b/fat-model-dependencies/pom.xml
@@ -221,5 +221,10 @@
       <artifactId>jdisc_http_service</artifactId>
       <version>${project.version}</version>
     </dependency>
+    <dependency>
+      <groupId>com.google.protobuf</groupId>
+      <artifactId>protobuf-java</artifactId>
+      <version>${protobuf.version}</version>
+    </dependency>
   </dependencies>
 </project>