diff --git a/tensorflow/lite/testing/op_tests/abs.py b/tensorflow/lite/testing/op_tests/abs.py
index 581831db120..6122d9e1f81 100644
--- a/tensorflow/lite/testing/op_tests/abs.py
+++ b/tensorflow/lite/testing/op_tests/abs.py
@@ -32,6 +32,7 @@ def make_abs_tests(options):
   test_parameters = [{
       "input_shape": [[], [1], [2, 3], [1, 1, 1, 1], [1, 3, 4, 3],
                       [3, 15, 14, 3], [3, 1, 2, 4, 6], [2, 2, 3, 4, 5, 6]],
+      "dynamic_range_quantize": [False, True]
   }]
 
   def build_graph(parameters):
diff --git a/tensorflow/lite/testing/toco_convert.py b/tensorflow/lite/testing/toco_convert.py
index f244fde55e3..4db2cf32fb1 100644
--- a/tensorflow/lite/testing/toco_convert.py
+++ b/tensorflow/lite/testing/toco_convert.py
@@ -103,10 +103,9 @@ def toco_convert(options, graph_def, input_tensors, output_tensors, **kwargs):
   input_arrays = [x[0] for x in input_tensors]
   data_types = [zip_test_utils.TF_TYPE_INFO[x[2]][1] for x in input_tensors]
 
-  if test_params.get("fully_quantize", False):
-    # Read the input range for the representative dataset from parameters.
-    min_value, max_value = test_params.get("input_range", (-1, 1))
-
+  fully_quantize = test_params.get("fully_quantize", False)
+  dynamic_range_quantize = test_params.get("dynamic_range_quantize", False)
+  if dynamic_range_quantize or fully_quantize:
     with tempfile.NamedTemporaryFile() as graphdef_file:
       graphdef_file.write(graph_def_str)
       graphdef_file.flush()
@@ -115,32 +114,38 @@ def toco_convert(options, graph_def, input_tensors, output_tensors, **kwargs):
       converter = tf.compat.v1.lite.TFLiteConverter.from_frozen_graph(
           graphdef_file.name, input_arrays, output_tensors, input_shapes)
 
-      def representative_dataset(input_tensors):
-        calibration_inputs = []
-        for _, shape, _ in input_tensors:
-          if shape:
-            dims = [dim.value for dim in shape.dims]
-            calibration_inputs.append(
-                np.random.uniform(min_value, max_value,
-                                  tuple(dims)).astype(np.float32))
-        return calibration_inputs
+      converter.optimizations = [tf.lite.Optimize.DEFAULT]
 
-      def representative_dataset_gen():
-        for _ in range(100):
-          yield representative_dataset(input_tensors)
+      if fully_quantize:
+        # Read the input range for the representative dataset from parameters.
+        min_value, max_value = test_params.get("input_range", (-1, 1))
 
-      converter.target_spec.supported_ops = [
-          tf.lite.OpsSet.TFLITE_BUILTINS_INT8
-      ]
-      converter.representative_dataset = representative_dataset_gen
-      if extra_toco_options.inference_input_type:
-        converter.inference_input_type = (
-            extra_toco_options.inference_input_type)
-      if extra_toco_options.inference_output_type:
-        converter.inference_output_type = (
-            extra_toco_options.inference_output_type)
-      else:
-        converter.inference_output_type = tf.int8
+        def representative_dataset(input_tensors):
+          calibration_inputs = []
+          for _, shape, _ in input_tensors:
+            if shape:
+              dims = [dim.value for dim in shape.dims]
+              calibration_inputs.append(
+                  np.random.uniform(min_value, max_value,
+                                    tuple(dims)).astype(np.float32))
+          return calibration_inputs
+
+        def representative_dataset_gen():
+          for _ in range(100):
+            yield representative_dataset(input_tensors)
+
+        converter.target_spec.supported_ops = [
+            tf.lite.OpsSet.TFLITE_BUILTINS_INT8
+        ]
+        converter.representative_dataset = representative_dataset_gen
+        if extra_toco_options.inference_input_type:
+          converter.inference_input_type = (
+              extra_toco_options.inference_input_type)
+        if extra_toco_options.inference_output_type:
+          converter.inference_output_type = (
+              extra_toco_options.inference_output_type)
+        else:
+          converter.inference_output_type = tf.int8
 
       try:
         tflite_model = converter.convert()