diff --git a/.azure-pipelines/linux-conda-CI.yml b/.azure-pipelines/linux-conda-CI.yml
index 25aa1d254..70fac27c0 100644
--- a/.azure-pipelines/linux-conda-CI.yml
+++ b/.azure-pipelines/linux-conda-CI.yml
@@ -19,7 +19,7 @@ jobs:
         python.version: '3.11'
         numpy.version: '>=1.21.1'
         scipy.version: '>=1.7.0'
-        onnx.version: '-i https://test.pypi.org/simple/ onnx==1.15.0rc2'
+        onnx.version: 'onnx==1.15.0'  # -i https://test.pypi.org/simple/ onnx==1.15.0rc2'
         onnx.target_opset: ''
         onnxrt.version: 'onnxruntime==1.16.1'
         sklearn.version: '>=1.3.1'
@@ -275,24 +275,40 @@ jobs:
       fi
     displayName: 'install onnx'
 
+  - script: |
+      pip show onnx
+    displayName: 'onnx version'
+
   - script: |
       pip install $(onnxrt.version)
     displayName: 'install onnxruntime'
 
+  - script: |
+      pip show onnx
+    displayName: 'onnx version'
+
   - script: |
       if [ '$(onnxcc.version)' == 'git' ]
       then
-        pip install git+https://github.com/microsoft/onnxconverter-common
+        pip install git+https://github.com/microsoft/onnxconverter-common --no-deps
       else
-        pip install onnxconverter-common$(onnxcc.version)
+        pip install onnxconverter-common$(onnxcc.version) --no-deps
       fi
     displayName: 'install onnxconverter-common'
 
+  - script: |
+      pip show onnx
+    displayName: 'onnx version'
+
   - script: |
       pip install -r requirements.txt
       pip install -r requirements-dev.txt
     displayName: 'install requirements'
 
+  - script: |
+      pip show onnx
+    displayName: 'onnx version'
+
   - script: |
       pip install -e .
     displayName: 'install'
diff --git a/.azure-pipelines/win32-conda-CI.yml b/.azure-pipelines/win32-conda-CI.yml
index 1fa00bda9..ea5a2c1c5 100644
--- a/.azure-pipelines/win32-conda-CI.yml
+++ b/.azure-pipelines/win32-conda-CI.yml
@@ -15,7 +15,7 @@ jobs:
     matrix:
       Py310-Onnx150-Rt161-Skl131:
         python.version: '3.11'
-        onnx.version: '-i https://test.pypi.org/simple/ onnx==1.15.0rc2'
+        onnx.version: 'onnx==1.15.0'  # '-i https://test.pypi.org/simple/ onnx==1.15.0rc2'
         onnx.target_opset: ''
         numpy.version: 'numpy>=1.22.3'
         scipy.version: 'scipy'
@@ -181,7 +181,7 @@ jobs:
 
   - script: |
       call activate skl2onnxEnvironment
-      pip install $(onnxcc.version)
+      pip install $(onnxcc.version) --no-deps
     displayName: 'install onnxconverter-common'
 
   - script: |
diff --git a/skl2onnx/operator_converters/gradient_boosting.py b/skl2onnx/operator_converters/gradient_boosting.py
index 42b598609..13a8a3597 100644
--- a/skl2onnx/operator_converters/gradient_boosting.py
+++ b/skl2onnx/operator_converters/gradient_boosting.py
@@ -39,7 +39,10 @@ def convert_sklearn_gradient_boosting_classifier(
     transform = "LOGISTIC" if op.n_classes_ == 2 else "SOFTMAX"
     if op.init == "zero":
         loss = op._loss if hasattr(op, "_loss") else op.loss_
-        base_values = np.zeros(loss.K)
+        if hasattr(loss, "K"):
+            base_values = np.zeros(loss.K)
+        else:
+            base_values = np.zeros(1)
     elif op.init is None:
         if hasattr(op.estimators_[0, 0], "n_features_in_"):
             # sklearn >= 1.2
@@ -142,7 +145,10 @@ def convert_sklearn_gradient_boosting_regressor(
 
     if op.init == "zero":
         loss = op._loss if hasattr(op, "_loss") else op.loss_
-        cst = np.zeros(loss.K)
+        if hasattr(loss, "K"):
+            cst = np.zeros(loss.K)
+        else:
+            cst = np.zeros(1)
     elif op.init is None:
         # constant_ was introduced in scikit-learn 0.21.
         if hasattr(op.init_, "constant_"):
diff --git a/tests/test_sklearn_bagging_converter.py b/tests/test_sklearn_bagging_converter.py
index 69fe498cc..fe5df9fb1 100644
--- a/tests/test_sklearn_bagging_converter.py
+++ b/tests/test_sklearn_bagging_converter.py
@@ -3,6 +3,7 @@
 
 import unittest
 import packaging.version as pv
+import onnx
 import onnxruntime
 
 try:
@@ -351,6 +352,10 @@ def test_bagging_regressor_sgd(self):
             X, model, model_onnx, basename="SklearnBaggingRegressorSGD-Dec4"
         )
 
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="Fixed issue in more recent versions",
+    )
     @ignore_warnings(category=FutureWarning)
     def test_bagging_regressor_gradient_boosting(self):
         model, X = fit_regression_model(
diff --git a/tests/test_sklearn_count_vectorizer_converter.py b/tests/test_sklearn_count_vectorizer_converter.py
index cf21218a4..65bf03868 100644
--- a/tests/test_sklearn_count_vectorizer_converter.py
+++ b/tests/test_sklearn_count_vectorizer_converter.py
@@ -5,7 +5,9 @@
 """
 import unittest
 import sys
+import packaging.version as pv
 import numpy
+import onnx
 from sklearn.feature_extraction.text import CountVectorizer
 from skl2onnx import convert_sklearn
 from skl2onnx.common.data_types import StringTensorType
@@ -14,6 +16,10 @@
 
 class TestSklearnCountVectorizer(unittest.TestCase):
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     def test_model_count_vectorizer11(self):
         corpus = numpy.array(
             [
@@ -36,6 +42,10 @@ def test_model_count_vectorizer11(self):
             corpus, vect, model_onnx, basename="SklearnCountVectorizer11-OneOff-SklCol"
         )
 
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
     def test_model_count_vectorizer22(self):
         corpus = numpy.array(
@@ -59,6 +69,10 @@ def test_model_count_vectorizer22(self):
             corpus, vect, model_onnx, basename="SklearnCountVectorizer22-OneOff-SklCol"
         )
 
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
     def test_model_count_vectorizer12(self):
         corpus = numpy.array(
@@ -82,6 +96,10 @@ def test_model_count_vectorizer12(self):
             corpus, vect, model_onnx, basename="SklearnCountVectorizer12-OneOff-SklCol"
         )
 
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
     def test_model_count_vectorizer13(self):
         corpus = numpy.array(
@@ -105,6 +123,10 @@ def test_model_count_vectorizer13(self):
             corpus, vect, model_onnx, basename="SklearnCountVectorizer13-OneOff-SklCol"
         )
 
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
     def test_model_count_vectorizer_binary(self):
         corpus = numpy.array(
@@ -131,6 +153,10 @@ def test_model_count_vectorizer_binary(self):
             basename="SklearnCountVectorizerBinary-OneOff-SklCol",
         )
 
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
     def test_model_count_vectorizer11_locale(self):
         corpus = numpy.array(
diff --git a/tests/test_sklearn_count_vectorizer_converter_bug.py b/tests/test_sklearn_count_vectorizer_converter_bug.py
index 73b4fea1d..aaea67226 100644
--- a/tests/test_sklearn_count_vectorizer_converter_bug.py
+++ b/tests/test_sklearn_count_vectorizer_converter_bug.py
@@ -4,7 +4,9 @@
 Tests scikit-learn's count vectorizer converter.
 """
 import unittest
+import packaging.version as pv
 import numpy
+import onnx
 from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer
 from skl2onnx import convert_sklearn
 from skl2onnx.common.data_types import StringTensorType
@@ -12,6 +14,10 @@
 
 
 class TestSklearnCountVectorizerBug(unittest.TestCase):
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
     def test_model_count_vectorizer_custom_tokenizer(self):
         corpus = numpy.array(
@@ -47,6 +53,10 @@ def test_model_count_vectorizer_custom_tokenizer(self):
             basename="SklearnTfidfVectorizer11CustomTokenizer-OneOff-SklCol",
         )
 
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
     def test_model_count_vectorizer_wrong_ngram(self):
         corpus = numpy.array(
diff --git a/tests/test_sklearn_documentation.py b/tests/test_sklearn_documentation.py
index ce9e5922b..00acc45f7 100644
--- a/tests/test_sklearn_documentation.py
+++ b/tests/test_sklearn_documentation.py
@@ -7,7 +7,9 @@
 import unittest
 import urllib.error
 import warnings
+import packaging.version as pv
 import numpy as np
+import onnx
 
 from sklearn.base import BaseEstimator, TransformerMixin
 from sklearn.datasets import fetch_20newsgroups
@@ -73,6 +75,10 @@ class TestSklearnDocumentation(unittest.TestCase):
     @unittest.skipIf(
         TARGET_OPSET < 10, reason="Encoding issue fixed in a later version"
     )
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     def test_pipeline_tfidf(self):
         categories = ["alt.atheism", "talk.religion.misc"]
         try:
@@ -112,6 +118,10 @@ def test_pipeline_tfidf(self):
     @unittest.skipIf(
         TARGET_OPSET < 10, reason="Encoding issue fixed in a later version"
     )
+    @unittest.skipIf(
+        pv.Version(onnx.__version__) < pv.Version("1.16.0"),
+        reason="ReferenceEvaluator does not support tfidf with strings",
+    )
     def test_pipeline_tfidf_pipeline_minmax(self):
         categories = ["alt.atheism", "talk.religion.misc"]
         try:
diff --git a/tests/test_sklearn_glm_regressor_converter.py b/tests/test_sklearn_glm_regressor_converter.py
index 063fa43fd..61259a8ff 100644
--- a/tests/test_sklearn_glm_regressor_converter.py
+++ b/tests/test_sklearn_glm_regressor_converter.py
@@ -4,6 +4,7 @@
 
 import unittest
 import packaging.version as pv
+import onnx
 import numpy
 from numpy.testing import assert_almost_equal
 
@@ -54,6 +55,12 @@
 ort_version = ort_version.split("+")[0]
 skl_version = ".".join(sklearn_version.split(".")[:2])
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
 
 class TestGLMRegressorConverter(unittest.TestCase):
     @ignore_warnings(category=(FutureWarning, ConvergenceWarning))
@@ -778,6 +785,7 @@ def test_model_ransac_regressor_tree(self):
             model_onnx,
             verbose=False,
             basename="SklearnRANSACRegressorTree-Dec3",
+            backend=BACKEND,
         )
 
     @ignore_warnings(category=(FutureWarning, ConvergenceWarning))
diff --git a/tests/test_sklearn_gradient_boosting_converters.py b/tests/test_sklearn_gradient_boosting_converters.py
index 4ff10f1b6..45197ba6c 100644
--- a/tests/test_sklearn_gradient_boosting_converters.py
+++ b/tests/test_sklearn_gradient_boosting_converters.py
@@ -1,9 +1,9 @@
 # SPDX-License-Identifier: Apache-2.0
 
-
 import packaging.version as pv
 import unittest
 import numpy as np
+import onnx
 from pandas import DataFrame
 from sklearn import __version__ as skl_version
 from sklearn.datasets import make_classification
@@ -30,6 +30,12 @@
 ort_version = ort_version.split("+")[0]
 skl_version = skl_version.split("+")[0]
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
 
 class TestSklearnGradientBoostingModels(unittest.TestCase):
     @unittest.skipIf(
@@ -215,7 +221,11 @@ def test_gradient_boosting_regressor_ls_loss(self):
         )
         self.assertIsNotNone(model_onnx)
         dump_data_and_model(
-            X, model, model_onnx, basename="SklearnGradientBoostingRegressionLsLoss"
+            X,
+            model,
+            model_onnx,
+            basename="SklearnGradientBoostingRegressionLsLoss",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(
@@ -233,7 +243,11 @@ def test_gradient_boosting_regressor_lad_loss(self):
         )
         self.assertIsNotNone(model_onnx)
         dump_data_and_model(
-            X, model, model_onnx, basename="SklearnGradientBoostingRegressionLadLoss"
+            X,
+            model,
+            model_onnx,
+            basename="SklearnGradientBoostingRegressionLadLoss",
+            backend=BACKEND,
         )
 
     def test_gradient_boosting_regressor_huber_loss(self):
@@ -248,7 +262,11 @@ def test_gradient_boosting_regressor_huber_loss(self):
         )
         self.assertIsNotNone(model_onnx)
         dump_data_and_model(
-            X, model, model_onnx, basename="SklearnGradientBoostingRegressionHuberLoss"
+            X,
+            model,
+            model_onnx,
+            basename="SklearnGradientBoostingRegressionHuberLoss",
+            backend=BACKEND,
         )
 
     def test_gradient_boosting_regressor_quantile_loss(self):
@@ -267,6 +285,7 @@ def test_gradient_boosting_regressor_quantile_loss(self):
             model,
             model_onnx,
             basename="SklearnGradientBoostingRegressionQuantileLoss-Dec4",
+            backend=BACKEND,
         )
 
     def test_gradient_boosting_regressor_int(self):
@@ -300,6 +319,7 @@ def test_gradient_boosting_regressor_zero_init(self):
             model,
             model_onnx,
             basename="SklearnGradientBoostingRegressionZeroInit-Dec4",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(
diff --git a/tests/test_sklearn_grid_search_cv_converter.py b/tests/test_sklearn_grid_search_cv_converter.py
index 1e29191d8..3144d57fc 100644
--- a/tests/test_sklearn_grid_search_cv_converter.py
+++ b/tests/test_sklearn_grid_search_cv_converter.py
@@ -61,10 +61,15 @@ def test_grid_search_multiclass_float(self):
             "GridSearchCV",
             [("input", FloatTensorType([None, X.shape[1]]))],
             target_opset=TARGET_OPSET,
+            options={"zipmap": False},
         )
         self.assertIsNotNone(model_onnx)
         dump_data_and_model(
-            X, model, model_onnx, basename="SklearnGridSearchMulticlassFloat"
+            X,
+            model,
+            model_onnx,
+            basename="SklearnGridSearchMulticlassFloat",
+            backend="onnxruntime",
         )
 
     def test_grid_search_binary_int(self):
diff --git a/tests/test_sklearn_nearest_neighbour_converter.py b/tests/test_sklearn_nearest_neighbour_converter.py
index 76d6b04d7..1160983aa 100644
--- a/tests/test_sklearn_nearest_neighbour_converter.py
+++ b/tests/test_sklearn_nearest_neighbour_converter.py
@@ -145,7 +145,7 @@ def test_model_knn_regressor(self):
 
     @unittest.skipIf(dont_test_radius(), reason="not available")
     @unittest.skipIf(
-        pv.Version(ort_version) < pv.Version("1.8.0"), reason="produces nan values"
+        pv.Version(ort_version) < pv.Version("1.15.0"), reason="produces nan values"
     )
     @ignore_warnings(category=DeprecationWarning)
     def test_model_knn_regressor_radius(self):
diff --git a/tests/test_sklearn_one_vs_rest_classifier_converter.py b/tests/test_sklearn_one_vs_rest_classifier_converter.py
index 10c6bafbd..ba0ba871a 100644
--- a/tests/test_sklearn_one_vs_rest_classifier_converter.py
+++ b/tests/test_sklearn_one_vs_rest_classifier_converter.py
@@ -409,7 +409,11 @@ def test_ovr_classification_float_binary_ensemble(self):
         )
         self.assertIsNotNone(model_onnx)
         dump_data_and_model(
-            X, model, model_onnx, basename="SklearnOVRClassificationFloatBinEnsemble"
+            X,
+            model,
+            model_onnx,
+            basename="SklearnOVRClassificationFloatBinEnsemble",
+            backend="onnxruntime",
         )
 
     @ignore_warnings(category=warnings_to_skip)
diff --git a/tests/test_sklearn_pipeline_within_pipeline.py b/tests/test_sklearn_pipeline_within_pipeline.py
index e45de2305..77e833d8f 100644
--- a/tests/test_sklearn_pipeline_within_pipeline.py
+++ b/tests/test_sklearn_pipeline_within_pipeline.py
@@ -6,7 +6,9 @@
 from textwrap import dedent
 import unittest
 from io import StringIO
+import packaging.version as pv
 import numpy as np
+import onnx
 import pandas
 
 try:
@@ -33,6 +35,12 @@
 from skl2onnx.common.data_types import FloatTensorType, StringTensorType
 from test_utils import dump_data_and_model, TARGET_OPSET
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
 
 class TestSklearnPipelineWithinPipeline(unittest.TestCase):
     def test_pipeline_pca_pipeline_minmax(self):
@@ -362,7 +370,11 @@ def test_complex_pipeline(self):
             preprocessor, initial_types=initial_type, target_opset=TARGET_OPSET
         )
         dump_data_and_model(
-            X_train, preprocessor, onx, basename="SklearnPipelineComplex"
+            X_train,
+            preprocessor,
+            onx,
+            basename="SklearnPipelineComplex",
+            backend=BACKEND,
         )
 
 
diff --git a/tests/test_sklearn_random_forest_converters.py b/tests/test_sklearn_random_forest_converters.py
index c4b2f9db2..d647dd166 100644
--- a/tests/test_sklearn_random_forest_converters.py
+++ b/tests/test_sklearn_random_forest_converters.py
@@ -5,6 +5,7 @@
 import packaging.version as pv
 import numpy
 from numpy.testing import assert_almost_equal
+import onnx
 from onnxruntime import InferenceSession, __version__ as ort_version
 import sklearn
 from sklearn.datasets import load_iris, make_regression, make_classification
@@ -66,6 +67,12 @@ def _sklearn_version():
 
 ort_version = ".".join(ort_version.split(".")[:2])
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
 
 class TestSklearnTreeEnsembleModels(unittest.TestCase):
     def test_random_forest_classifier(self):
@@ -367,6 +374,7 @@ def common_test_model_hgb_regressor(self, add_nan=False):
             model_onnx,
             basename=f"SklearnHGBRegressor{add_nan}",
             verbose=False,
+            backend=BACKEND,
         )
 
     @unittest.skipIf(
diff --git a/tests/test_sklearn_scaler_converter.py b/tests/test_sklearn_scaler_converter.py
index cde0231a5..94128bac0 100644
--- a/tests/test_sklearn_scaler_converter.py
+++ b/tests/test_sklearn_scaler_converter.py
@@ -50,6 +50,7 @@ def test_standard_scaler_int(self):
             model,
             model_onnx,
             basename="SklearnStandardScalerInt64",
+            backend="onnxruntime",
         )
 
     @ignore_warnings(category=DeprecationWarning)
diff --git a/tests/test_sklearn_text.py b/tests/test_sklearn_text.py
index ae8f33ea0..ebd966509 100644
--- a/tests/test_sklearn_text.py
+++ b/tests/test_sklearn_text.py
@@ -1,10 +1,10 @@
 # -*- coding: utf-8 -*-
-"""
-@brief      test log(time=10s)
-"""
+
 import unittest
+import packaging.version as pv
 import numpy
 from numpy.testing import assert_almost_equal
+import onnx
 from sklearn import __version__ as skl_version, __file__ as skl_file
 from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer
 from skl2onnx import to_onnx
@@ -13,6 +13,12 @@
 from skl2onnx.common.data_types import StringTensorType
 from test_utils import dump_data_and_model, TARGET_OPSET
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
 
 class TestSklearnText(unittest.TestCase):
     def test_count_vectorizer(self):
@@ -297,6 +303,7 @@ def test_model_tfidf_vectorizer_issue(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizerIssue-OneOff-SklCol",
+            backend=BACKEND,
         )
 
 
diff --git a/tests/test_sklearn_tfidf_transformer_converter_sparse.py b/tests/test_sklearn_tfidf_transformer_converter_sparse.py
index cb3353a7b..19c2e1831 100644
--- a/tests/test_sklearn_tfidf_transformer_converter_sparse.py
+++ b/tests/test_sklearn_tfidf_transformer_converter_sparse.py
@@ -6,6 +6,7 @@
 """
 import packaging.version as pv
 import unittest
+import onnx
 from sklearn.datasets import fetch_20newsgroups
 from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer
 from sklearn.pipeline import Pipeline
@@ -14,6 +15,12 @@
 from skl2onnx import convert_sklearn
 from test_utils import dump_data_and_model, TARGET_OPSET
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
 
 class TestSklearnTfidfVectorizerSparse(unittest.TestCase):
     @unittest.skipIf(
@@ -51,6 +58,7 @@ def test_model_tfidf_transform_bug(self):
             text_clf,
             model_onnx,
             basename="SklearnPipelineTfidfTransformer",
+            backend=BACKEND,
         )
 
 
diff --git a/tests/test_sklearn_tfidf_vectorizer_converter.py b/tests/test_sklearn_tfidf_vectorizer_converter.py
index f72f9080f..117286488 100644
--- a/tests/test_sklearn_tfidf_vectorizer_converter.py
+++ b/tests/test_sklearn_tfidf_vectorizer_converter.py
@@ -9,6 +9,7 @@
 import packaging.version as pv
 import numpy
 from numpy.testing import assert_almost_equal
+import onnx
 from sklearn.feature_extraction.text import TfidfVectorizer
 from sklearn.svm import SVC
 
@@ -34,6 +35,12 @@
 
 ort_version = ".".join(ort_version.split(".")[:2])
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
 
 class TestSklearnTfidfVectorizer(unittest.TestCase):
     def get_options(self):
@@ -63,7 +70,11 @@ def test_model_tfidf_vectorizer11(self):
         )
         self.assertTrue(model_onnx is not None)
         dump_data_and_model(
-            corpus, vect, model_onnx, basename="SklearnTfidfVectorizer11-OneOff-SklCol"
+            corpus,
+            vect,
+            model_onnx,
+            basename="SklearnTfidfVectorizer11-OneOff-SklCol",
+            backend=BACKEND,
         )
 
         sess = InferenceSession(
@@ -100,6 +111,7 @@ def test_model_tfidf_vectorizer11_nolowercase(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11NoL-OneOff-SklCol",
+            backend=BACKEND,
         )
 
         sess = InferenceSession(
@@ -170,6 +182,7 @@ def test_model_tfidf_vectorizer11_empty_string_case1(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11EmptyStringSepCase1-" "OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -198,6 +211,7 @@ def test_model_tfidf_vectorizer11_empty_string_case2(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11EmptyString-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -233,6 +247,7 @@ def test_model_tfidf_vectorizer11_out_vocabulary(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11OutVocab-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -256,7 +271,11 @@ def test_model_tfidf_vectorizer22(self):
         )
         self.assertTrue(model_onnx is not None)
         dump_data_and_model(
-            corpus, vect, model_onnx, basename="SklearnTfidfVectorizer22-OneOff-SklCol"
+            corpus,
+            vect,
+            model_onnx,
+            basename="SklearnTfidfVectorizer22-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -273,7 +292,11 @@ def test_model_tfidf_vectorizer21(self):
         )
         self.assertTrue(model_onnx is not None)
         dump_data_and_model(
-            corpus, vect, model_onnx, basename="SklearnTfidfVectorizer22S-OneOff-SklCol"
+            corpus,
+            vect,
+            model_onnx,
+            basename="SklearnTfidfVectorizer22S-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -297,7 +320,11 @@ def test_model_tfidf_vectorizer12(self):
         )
         self.assertTrue(model_onnx is not None)
         dump_data_and_model(
-            corpus, vect, model_onnx, basename="SklearnTfidfVectorizer22-OneOff-SklCol"
+            corpus,
+            vect,
+            model_onnx,
+            basename="SklearnTfidfVectorizer22-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -324,6 +351,7 @@ def test_model_tfidf_vectorizer12_normL1(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer22L1-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -351,6 +379,7 @@ def test_model_tfidf_vectorizer12_normL2(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer22L2-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -374,7 +403,11 @@ def test_model_tfidf_vectorizer13(self):
         )
         self.assertTrue(model_onnx is not None)
         dump_data_and_model(
-            corpus, vect, model_onnx, basename="SklearnTfidfVectorizer13-OneOff-SklCol"
+            corpus,
+            vect,
+            model_onnx,
+            basename="SklearnTfidfVectorizer13-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -409,6 +442,7 @@ def test_model_tfidf_vectorizer11parenthesis_class(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11ParenthesisClass-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -454,6 +488,7 @@ def test_model_tfidf_vectorizer11_idparenthesis_id(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11ParenthesisId-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -481,6 +516,7 @@ def test_model_tfidf_vectorizer_binary(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizerBinary-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -511,6 +547,7 @@ def test_model_tfidf_vectorizer11_64(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer1164-OneOff-SklCol",
+            backend=BACKEND,
         )
 
         sess = InferenceSession(
@@ -614,6 +651,7 @@ def test_model_tfidf_vectorizer11_custom_vocabulary(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11CustomVocab-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -650,6 +688,7 @@ def test_model_tfidf_vectorizer_locale(self):
                 vect,
                 model_onnx,
                 basename="SklearnTfidfVectorizer11Locale-OneOff-SklCol",
+                backend=BACKEND,
             )
 
             sess = InferenceSession(
diff --git a/tests/test_sklearn_tfidf_vectorizer_converter_char.py b/tests/test_sklearn_tfidf_vectorizer_converter_char.py
index f0096b18e..b5abf06b6 100644
--- a/tests/test_sklearn_tfidf_vectorizer_converter_char.py
+++ b/tests/test_sklearn_tfidf_vectorizer_converter_char.py
@@ -4,13 +4,22 @@
 Tests scikit-learn's tfidf converter.
 """
 import unittest
+import packaging.version as pv
 import numpy
+import onnx
 from sklearn.feature_extraction.text import TfidfVectorizer
 from skl2onnx import convert_sklearn
 from skl2onnx.common.data_types import StringTensorType
 from test_utils import dump_data_and_model, TARGET_OPSET
 
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
+
 class TestSklearnTfidfVectorizerRegex(unittest.TestCase):
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
     def test_model_tfidf_vectorizer11_short_word(self):
@@ -37,6 +46,7 @@ def test_model_tfidf_vectorizer11_short_word(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11CharW2-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -84,6 +94,7 @@ def test_model_tfidf_vectorizer11_char(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11Char-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -110,6 +121,7 @@ def test_model_tfidf_vectorizer11_char_doublespace(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11CharSpace-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -135,6 +147,7 @@ def test_model_tfidf_vectorizer12_char(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer12Char-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -161,6 +174,7 @@ def test_model_tfidf_vectorizer12_normL1_char(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer12L1Char-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -212,6 +226,7 @@ def test_model_tfidf_vectorizer11_short_word_spaces(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11CharW2-OneOff-SklCol",
+            backend=BACKEND,
         )
 
 
diff --git a/tests/test_sklearn_tfidf_vectorizer_converter_dataset.py b/tests/test_sklearn_tfidf_vectorizer_converter_dataset.py
index 9ed7306ae..779160251 100644
--- a/tests/test_sklearn_tfidf_vectorizer_converter_dataset.py
+++ b/tests/test_sklearn_tfidf_vectorizer_converter_dataset.py
@@ -4,7 +4,9 @@
 Tests scikit-learn's tfidf converter using downloaded data.
 """
 import unittest
+import packaging.version as pv
 import numpy as np
+import onnx
 from sklearn.model_selection import train_test_split
 from sklearn.feature_extraction.text import TfidfVectorizer
 from sklearn.datasets import fetch_20newsgroups
@@ -12,6 +14,12 @@
 from skl2onnx.common.data_types import StringTensorType
 from test_utils import dump_data_and_model, TARGET_OPSET
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
 
 class TestSklearnTfidfVectorizerDataSet(unittest.TestCase):
     @unittest.skipIf(TARGET_OPSET < 9, reason="not available")
@@ -30,7 +38,11 @@ def test_tfidf_20newsgroups(self):
             target_opset=TARGET_OPSET,
         )
         dump_data_and_model(
-            X_test, model, onnx_model, basename="SklearnTfidfVectorizer20newsgroups"
+            X_test,
+            model,
+            onnx_model,
+            basename="SklearnTfidfVectorizer20newsgroups",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 9, reason="not available")
@@ -53,6 +65,7 @@ def test_tfidf_20newsgroups_nolowercase(self):
             model,
             onnx_model,
             basename="SklearnTfidfVectorizer20newsgroupsNOLower",
+            backend=BACKEND,
         )
 
 
diff --git a/tests/test_sklearn_tfidf_vectorizer_converter_regex.py b/tests/test_sklearn_tfidf_vectorizer_converter_regex.py
index df94705f4..184fd7ceb 100644
--- a/tests/test_sklearn_tfidf_vectorizer_converter_regex.py
+++ b/tests/test_sklearn_tfidf_vectorizer_converter_regex.py
@@ -4,13 +4,22 @@
 Tests scikit-learn's tfidf converter.
 """
 import unittest
+import packaging.version as pv
 import numpy
+import onnx
 from sklearn.feature_extraction.text import TfidfVectorizer
 from skl2onnx import convert_sklearn
 from skl2onnx.common.data_types import StringTensorType
 from test_utils import dump_data_and_model, TARGET_OPSET
 
 
+BACKEND = (
+    "onnxruntime"
+    if pv.Version(onnx.__version__) < pv.Version("1.16.0")
+    else "onnx;onnxruntime"
+)
+
+
 class TestSklearnTfidfVectorizerRegex(unittest.TestCase):
     def get_options(self):
         return {TfidfVectorizer: {"tokenexp": ""}}
@@ -40,6 +49,7 @@ def test_model_tfidf_vectorizer11(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11Regex-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -70,7 +80,13 @@ def test_model_tfidf_vectorizer11_opset(self):
             self.assertTrue(model_onnx is not None)
             if opset >= 10:
                 name = "SklearnTfidfVectorizer11Rx%d-OneOff-SklCol" % opset
-                dump_data_and_model(corpus, vect, model_onnx, basename=name)
+                dump_data_and_model(
+                    corpus,
+                    vect,
+                    model_onnx,
+                    basename=name,
+                    backend=BACKEND,
+                )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
     def test_model_tfidf_vectorizer11_word4(self):
@@ -99,6 +115,7 @@ def test_model_tfidf_vectorizer11_word4(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11Regex4-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -127,6 +144,7 @@ def test_model_tfidf_vectorizer11_empty_string(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11EmptyStringRegex-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -162,6 +180,7 @@ def test_model_tfidf_vectorizer11_out_vocabulary(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11OutVocabRegex-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -189,6 +208,7 @@ def test_model_tfidf_vectorizer22(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer22Regex-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -214,6 +234,7 @@ def test_model_tfidf_vectorizer12(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer12SRegex-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -241,6 +262,7 @@ def test_model_tfidf_vectorizer122(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer12Regex-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -267,6 +289,7 @@ def test_model_tfidf_vectorizer12_normL1(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer12L1Regex-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -294,6 +317,7 @@ def test_model_tfidf_vectorizer12_normL2(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer12L2Regex-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -321,6 +345,7 @@ def test_model_tfidf_vectorizer13(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer13Regex-OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -356,6 +381,7 @@ def test_model_tfidf_vectorizer11parenthesis_class(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11ParenthesisClassRegex-" "OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -409,6 +435,7 @@ def test_model_tfidf_vectorizer11_idparenthesis_id(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizer11ParenthesisIdRegex-" "OneOff-SklCol",
+            backend=BACKEND,
         )
 
     @unittest.skipIf(TARGET_OPSET < 10, reason="not available")
@@ -457,9 +484,9 @@ def test_model_tfidf_vectorizer_issue(self):
             vect,
             model_onnx,
             basename="SklearnTfidfVectorizerIssue-OneOff-SklCol",
+            backend=BACKEND,
         )
 
 
 if __name__ == "__main__":
-    TestSklearnTfidfVectorizerRegex().test_model_tfidf_vectorizer_issue()
-    unittest.main()
+    unittest.main(verbosity=2)
diff --git a/tests/test_utils/__init__.py b/tests/test_utils/__init__.py
index 6bc2a07e9..bb9385e7e 100644
--- a/tests/test_utils/__init__.py
+++ b/tests/test_utils/__init__.py
@@ -54,10 +54,9 @@ def InferenceSessionEx(onx, *args, verbose=0, **kwargs):
 def create_tensor(N, C, H=None, W=None):
     if H is None and W is None:
         return np.random.rand(N, C).astype(np.float32, copy=False)
-    elif H is not None and W is not None:
+    if H is not None and W is not None:
         return np.random.rand(N, C, H, W).astype(np.float32, copy=False)
-    else:
-        raise ValueError("This function only produce 2-D or 4-D tensor.")
+    raise ValueError("This function only produce 2-D or 4-D tensor.")
 
 
 def _get_ir_version(opv):
diff --git a/tests_onnxmltools/test_xgboost_converters.py b/tests_onnxmltools/test_xgboost_converters.py
index 404796a52..9e48209c6 100644
--- a/tests_onnxmltools/test_xgboost_converters.py
+++ b/tests_onnxmltools/test_xgboost_converters.py
@@ -74,7 +74,7 @@ def custom_parser(scope, model, inputs, custom_parsers=None):
         )
 
     @unittest.skipIf(
-        pv.Version(onnxmltools.__version__) < pv.Version("1.11"),
+        pv.Version(onnxmltools.__version__) < pv.Version("1.12"),
         reason="converter for xgboost is too old",
     )
     def test_xgb_regressor(self):