From 2a4e18f6a8510d47bc903f37cc50a0b2d304255e Mon Sep 17 00:00:00 2001
From: Lester Solbakken <lesters@oath.com>
Date: Wed, 27 Jan 2021 15:27:50 +0100
Subject: Propagate type of transformer token helper functions

---
 .../searchdefinition/MapEvaluationTypeContext.java   | 20 ++++++++++++++++++++
 .../expressiontransforms/TokenTransformer.java       |  2 +-
 .../src/test/integration/onnx-model/schemas/test.sd  | 10 +++++++++-
 .../RankingExpressionWithOnnxModelTestCase.java      | 14 +++++++++-----
 4 files changed, 39 insertions(+), 7 deletions(-)

(limited to 'config-model')

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 9dcba72161b..bc98f0ab8c5 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/MapEvaluationTypeContext.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/MapEvaluationTypeContext.java
@@ -2,6 +2,7 @@
 package com.yahoo.searchdefinition;
 
 import com.yahoo.searchdefinition.expressiontransforms.OnnxModelTransformer;
+import com.yahoo.searchdefinition.expressiontransforms.TokenTransformer;
 import com.yahoo.searchlib.rankingexpression.ExpressionFunction;
 import com.yahoo.searchlib.rankingexpression.RankingExpression;
 import com.yahoo.searchlib.rankingexpression.Reference;
@@ -165,6 +166,12 @@ public class MapEvaluationTypeContext extends FunctionReferenceContext implement
                 return onnxFeatureType.get();
             }
 
+            // A reference to a feature for transformer token input?
+            Optional<TensorType> transformerTokensFeatureType = transformerTokensFeatureType(reference);
+            if (transformerTokensFeatureType.isPresent()) {
+                return transformerTokensFeatureType.get();
+            }
+
             // A reference to a feature which returns a tensor?
             Optional<TensorType> featureTensorType = tensorFeatureType(reference);
             if (featureTensorType.isPresent()) {
@@ -237,6 +244,19 @@ public class MapEvaluationTypeContext extends FunctionReferenceContext implement
         return Optional.of(featureTypes.get(reference));
     }
 
+    private Optional<TensorType> transformerTokensFeatureType(Reference reference) {
+        if ( ! reference.name().equals("tokenTypeIds") &&
+                ! reference.name().equals("tokenInputIds") &&
+                ! reference.name().equals("tokenAttentionMask"))
+            return Optional.empty();
+
+        if ( ! (reference.arguments().size() > 1))
+            throw new IllegalArgumentException(reference.name() + " must have at least 2 arguments");
+
+        ExpressionNode size = reference.arguments().expressions().get(0);
+        return Optional.of(TokenTransformer.createTensorType(reference.name(), size));
+    }
+
     /**
      * 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/expressiontransforms/TokenTransformer.java b/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/TokenTransformer.java
index 1bb38eda9ff..192fb9baa9a 100644
--- a/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/TokenTransformer.java
+++ b/config-model/src/main/java/com/yahoo/searchdefinition/expressiontransforms/TokenTransformer.java
@@ -186,7 +186,7 @@ public class TokenTransformer extends ExpressionTransformer<RankProfileTransform
         }
     }
 
-    private TensorType createTensorType(String featureName, ExpressionNode argument) {
+    public static TensorType createTensorType(String featureName, ExpressionNode argument) {
         try {
             int length = Integer.parseInt(argument.toString());
             return new TensorType.Builder(TensorType.Value.FLOAT).indexed("d0", 1).indexed("d1", length).build();
diff --git a/config-model/src/test/integration/onnx-model/schemas/test.sd b/config-model/src/test/integration/onnx-model/schemas/test.sd
index eac6f650fef..5b440e80bed 100644
--- a/config-model/src/test/integration/onnx-model/schemas/test.sd
+++ b/config-model/src/test/integration/onnx-model/schemas/test.sd
@@ -99,6 +99,15 @@ search test {
         }
     }
 
+    rank-profile test_dynamic_model_with_transformer_tokens {
+        function my_function() {
+            expression: tokenTypeIds(2, attribute(document_field))
+        }
+        first-phase {
+            expression: onnx(dynamic_model){d0:0,d1:1}
+        }
+    }
+
     rank-profile test_unbound_model {
         function my_function() {
             expression: tensor(d0[1],d1[2])(d1)
@@ -108,5 +117,4 @@ search test {
         }
     }
 
-
 }
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 f8a379b4027..73ff4ac3bcd 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
@@ -113,7 +113,7 @@ public class RankingExpressionWithOnnxModelTestCase {
         RankProfilesConfig.Builder builder = new RankProfilesConfig.Builder();
         ((RankProfilesConfig.Producer) db).getConfig(builder);
         RankProfilesConfig config = new RankProfilesConfig(builder);
-        assertEquals(8, config.rankprofile().size());
+        assertEquals(9, config.rankprofile().size());
 
         assertEquals("test_model_config", config.rankprofile(2).name());
         assertEquals("rankingExpression(my_function).rankingScript", config.rankprofile(2).fef().property(0).name());
@@ -150,10 +150,14 @@ public class RankingExpressionWithOnnxModelTestCase {
         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());
 
-        assertEquals("test_unbound_model", config.rankprofile(7).name());
-        assertEquals("rankingExpression(my_function).rankingScript", config.rankprofile(7).fef().property(0).name());
-        assertEquals("rankingExpression(firstphase).rankingScript", config.rankprofile(7).fef().property(3).name());
-        assertEquals("onnxModel(unbound_model).my_output{d0:0, d1:1}", config.rankprofile(7).fef().property(3).value());
+        assertEquals("test_dynamic_model_with_transformer_tokens", config.rankprofile(7).name());
+        assertEquals("rankingExpression(my_function).rankingScript", config.rankprofile(7).fef().property(1).name());
+        assertEquals("tensor<float>(d0[1],d1[2])((if (d1 < rankingExpression(__token_length@-1993461420) + 2, 0, 1)))", config.rankprofile(7).fef().property(1).value());
+
+        assertEquals("test_unbound_model", config.rankprofile(8).name());
+        assertEquals("rankingExpression(my_function).rankingScript", config.rankprofile(8).fef().property(0).name());
+        assertEquals("rankingExpression(firstphase).rankingScript", config.rankprofile(8).fef().property(3).name());
+        assertEquals("onnxModel(unbound_model).my_output{d0:0, d1:1}", config.rankprofile(8).fef().property(3).value());
 
 
     }
-- 
cgit v1.2.3