Reduces greater batch size to 10.000 instead of 100.000.

_doc/examples/plot_op_onnx_topk.py

@@ -201,7 +201,7 @@ def f2(x, k=k): return fct2(x, k=k)
 
 
 K = [1, 2, 5, 10, 15]
-N = [1, 2, 3, 10, 100, 1000, 10000, 100000]
+N = [1, 2, 3, 10, 100, 1000, 10000]
 
 bench_para = benchmark_test(
     X32, (lambda x, k: topk_sorted_implementation_cpp(

_doc/examples/plot_opml_linear_regression.py

@@ -259,7 +259,7 @@ def autolabel(ax, rects):
 
 @ignore_warnings(category=FutureWarning)
 def run_bench(repeat=250, verbose=False):
-    n_obs = [1, 10, 100, 1000, 10000, 100000]
+    n_obs = [1, 10, 100, 1000, 10000]
     methods = ['predict']
     n_features = [10, 50]
     n_jobss = [cpu_count()]

_doc/examples/plot_opml_random_forest_cls_multi.py

@@ -311,7 +311,7 @@ def autolabel(ax, rects):
 
 @ignore_warnings(category=FutureWarning)
 def run_bench(repeat=100, verbose=False):
-    n_obs = [1, 10, 100, 1000, 10000, 100000]
+    n_obs = [1, 10, 100, 1000, 10000]
     methods = ['predict']
     n_features = [30]
     max_depths = [6, 8, 10, 12]

_doc/examples/plot_opml_random_forest_reg.py

@@ -335,7 +335,7 @@ def autolabel(ax, rects):
 
 @ignore_warnings(category=FutureWarning)
 def run_bench(repeat=100, verbose=False):
-    n_obs = [1, 10, 100, 1000, 10000, 100000]
+    n_obs = [1, 10, 100, 1000, 10000]
     methods = ['predict']
     n_features = [30]
     max_depths = [6, 8, 10, 12]

_doc/examples/plot_parallelism.py

@@ -42,7 +42,7 @@
 # +++++++++++++++++++++++++++++++
 
 
-data = make_regression(100000, 20)
+data = make_regression(50000, 20)
 X, y = data
 X_train, X_test, y_train, y_test = train_test_split(X, y)
 

_doc/sphinxdoc/source/_exts/generate_automated_pages.py

@@ -118,7 +118,7 @@ def write_page_onnxrt_benches(app, runtime, skip=None, white_list=None):
     opset_cols = [oc[1] for oc in opset_cols]
     new_cols = opset_cols[:1]
     bench_cols = ["RT/SKL-N=1", "N=10", "N=100",
-                  "N=1000", "N=10000", "N=100000"]
+                  "N=1000", "N=10000"]
     new_cols.extend(["ERROR-msg", "name", "problem", "scenario", 'optim'])
     new_cols.extend(bench_cols)
     new_cols.extend(opset_cols[1:])

_doc/sphinxdoc/source/onnx_bench.rst

@@ -155,7 +155,7 @@ it is *1/r* faster than *scikit-learn*.
             row["name"], row["problem"], row["scenario"], row["optim"],
             row["n_features"], row["opset"]).replace("-default|", "-*]"), axis=1)
     indices = ['label']
-    values = ['RT/SKL-N=1', 'N=10', 'N=100', 'N=1000', 'N=10000', 'N=100000']
+    values = ['RT/SKL-N=1', 'N=10', 'N=100', 'N=1000', 'N=10000']
     df1 = df1[indices + values]
     df2 = df2[indices + values]
     df = df1.merge(df2, on="label", suffixes=("__pyrtc", "__ort"), how='outer')

_doc/sphinxdoc/source/tutorial/benchmarkorts.rst

@@ -87,8 +87,8 @@ and with :epkg:`scikit-learn`.
     keep = [
         'name', 'problem', 'scenario', 'optim', 'n_features', 'runtime',
         'skl_version', 'opset11',
-        'RT/SKL-N=1', 'N=100000',
-        'RT/SKL-N=1-base', 'N=100000-base',
+        'RT/SKL-N=1',
+        'RT/SKL-N=1-base',
         ]
     suma = suma[keep].copy()
 
@@ -100,11 +100,11 @@ and with :epkg:`scikit-learn`.
             "'raw_scores': True", "RAW")
 
     suma['ORT ?x SKL ONE'] = 1. / suma["RT/SKL-N=1"]
-    suma['ORT ?x SKL BATCH'] = 1. / suma["N=100000"]
+    suma['ORT ?x SKL BATCH'] = 1. / suma["N=10000"]
     suma['NEW ?x ORT ONE'] = 1. / suma["RT/SKL-N=1-base"]
-    suma['NEW ?x ORT BATCH'] = 1. / suma["N=100000-base"]
+    suma['NEW ?x ORT BATCH'] = 1. / suma["N=10000-base"]
     suma['optim'] = suma['optim'].apply(replace)
-    suma = suma.drop(['RT/SKL-N=1', 'N=100000', 'RT/SKL-N=1-base', 'N=100000-base'], axis=1)
+    suma = suma.drop(['RT/SKL-N=1', 'N=10000', 'RT/SKL-N=1-base', 'N=10000-base'], axis=1)
 
     writer = pandas.ExcelWriter('merged.xlsx', engine='xlsxwriter')
     suma.to_excel(writer, index=False, float_format="%1.3f",

_unittests/ut_onnxrt/test_benchmark_replay.py

@@ -26,7 +26,7 @@ def test_benchmark_replay(self):
         rows = list(enumerate_benchmark_replay(
             temp, runtime='python', verbose=0))
         df = pandas.DataFrame(rows)
-        self.assertEqual(df.shape, (4, 39))
+        self.assertEqual(df.shape, (4, 35))
         self.assertIn('1000-skl-details', df.columns)
         self.assertIn('1000-skl', df.columns)
 
@@ -45,7 +45,7 @@ def test_benchmark_replay_onnxruntime(self):
         rows = list(enumerate_benchmark_replay(
             temp, runtime='onnxruntime', verbose=0))
         df = pandas.DataFrame(rows)
-        self.assertEqual(df.shape, (4, 39))
+        self.assertEqual(df.shape, (4, 35))
         self.assertIn('1000-skl-details', df.columns)
         self.assertIn('1000-skl', df.columns)
 

mlprodict/asv_benchmark/common_asv_skl.py

@@ -46,7 +46,7 @@ class _CommonAsvSklBenchmark:
 
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 10000, 100000],  # values for N
+        [1, 10, 100, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ["float", "double"],  # values for dtype

mlprodict/asv_benchmark/create_asv.py

@@ -64,7 +64,7 @@ def create_asv_benchmark(
         location, opset_min=-1, opset_max=None,
         runtime=('scikit-learn', 'python_compiled'), models=None,
         skip_models=None, extended_list=True,
-        dims=(1, 10, 100, 10000, 100000),
+        dims=(1, 10, 100, 10000),
         n_features=(4, 20), dtype=None,
         verbose=0, fLOG=print, clean=True,
         conf_params=None, filter_exp=None,
@@ -540,7 +540,7 @@ def _make_simple_name(name):
         # n_features, N, runtimes
         rep = {
             "['skl', 'pyrtc', 'ort'],  # values for runtime": str(runtime),
-            "[1, 10, 100, 1000, 10000, 100000],  # values for N": str(dims),
+            "[1, 10, 100, 1000, 10000],  # values for N": str(dims),
             "[4, 20],  # values for nf": str(n_features),
             "[get_opset_number_from_onnx()],  # values for opset": str(opsets),
             "['float', 'double'],  # values for dtype":

mlprodict/asv_benchmark/template/skl_model_classifier.py

@@ -29,7 +29,7 @@ class TemplateBenchmarkClassifier(_CommonAsvSklBenchmarkClassifier):
     """
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 1000, 10000, 100000],  # values for N
+        [1, 10, 100, 1000, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ['float', 'double'],  # values for dtype

mlprodict/asv_benchmark/template/skl_model_classifier_raw_scores.py

@@ -29,7 +29,7 @@ class TemplateBenchmarkClassifierRawScore(_CommonAsvSklBenchmarkClassifierRawSco
     """
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 1000, 10000, 100000],  # values for N
+        [1, 10, 100, 1000, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ['float', 'double'],  # values for dtype

mlprodict/asv_benchmark/template/skl_model_clustering.py

@@ -29,7 +29,7 @@ class TemplateBenchmarkClustering(_CommonAsvSklBenchmarkClustering):
     """
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 1000, 10000, 100000],  # values for N
+        [1, 10, 100, 1000, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ['float', 'double'],  # values for dtype

mlprodict/asv_benchmark/template/skl_model_multi_classifier.py

@@ -29,7 +29,7 @@ class TemplateBenchmarkMultiClassifier(_CommonAsvSklBenchmarkMultiClassifier):
     """
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 1000, 10000, 100000],  # values for N
+        [1, 10, 100, 1000, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ['float', 'double'],  # values for dtype

mlprodict/asv_benchmark/template/skl_model_outlier.py

@@ -29,7 +29,7 @@ class TemplateBenchmarkOutlier(_CommonAsvSklBenchmarkOutlier):
     """
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 1000, 10000, 100000],  # values for N
+        [1, 10, 100, 1000, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ['float', 'double'],  # values for dtype

mlprodict/asv_benchmark/template/skl_model_regressor.py

@@ -29,7 +29,7 @@ class TemplateBenchmarkRegressor(_CommonAsvSklBenchmarkRegressor):
     """
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 1000, 10000, 100000],  # values for N
+        [1, 10, 100, 1000, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ['float', 'double'],  # values for dtype

mlprodict/asv_benchmark/template/skl_model_trainable_transform.py

@@ -29,7 +29,7 @@ class TemplateBenchmarkTrainableTransform(_CommonAsvSklBenchmarkTrainableTransfo
     """
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 1000, 10000, 100000],  # values for N
+        [1, 10, 100, 1000, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ['float', 'double'],  # values for dtype

mlprodict/asv_benchmark/template/skl_model_transform.py

@@ -29,7 +29,7 @@ class TemplateBenchmarkTransform(_CommonAsvSklBenchmarkTransform):
     """
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 1000, 10000, 100000],  # values for N
+        [1, 10, 100, 1000, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ['float', 'double'],  # values for dtype

mlprodict/asv_benchmark/template/skl_model_transform_positive.py

@@ -29,7 +29,7 @@ class TemplateBenchmarkTransformPositive(_CommonAsvSklBenchmarkTransformPositive
     """
     params = [
         ['skl', 'pyrtc', 'ort'],  # values for runtime
-        [1, 10, 100, 1000, 10000, 100000],  # values for N
+        [1, 10, 100, 1000, 10000],  # values for N
         [4, 20],  # values for nf
         [get_opset_number_from_onnx()],  # values for opset
         ['float', 'double'],  # values for dtype

mlprodict/cli/asv_bench.py

@@ -10,7 +10,7 @@
 def asv_bench(location='asvsklonnx', opset_min=-1, opset_max=None,
               runtime='scikit-learn,python_compiled', models=None,
               skip_models=None, extended_list=True,
-              dims='1,10,100,1000,10000,100000',
+              dims='1,10,100,1000,10000',
               n_features='4,20', dtype=None,
               verbose=1, fLOG=print, clean=True, flat=False,
               conf_params=None, build=None, add_pyspy=False,

mlprodict/onnx_conv/operator_converters/conv_lightgbm.py

@@ -110,8 +110,8 @@ def _parse_tree_structure(tree_id, class_id, learning_rate, tree_structure, attr
         except ValueError as e:  # pragma: no cover
             import pprint
             text = pprint.pformat(tree_structure)
-            if len(text) > 100000:
-                text = text[:100000] + "\n..."
+            if len(text) > 99999:
+                text = text[:99999] + "\n..."
             raise TypeError("threshold must be a number not '{}'"
                             "\n{}".format(tree_structure['threshold'], text)) from e
     else:
@@ -173,8 +173,8 @@ def _parse_node(tree_id, class_id, node_id, node_id_pool, node_pyid_pool,
             except ValueError as e:  # pragma: no cover
                 import pprint
                 text = pprint.pformat(node)
-                if len(text) > 100000:
-                    text = text[:100000] + "\n..."
+                if len(text) > 99999:
+                    text = text[:99999] + "\n..."
                 raise TypeError("threshold must be a number not '{}'"
                                 "\n{}".format(node['threshold'], text)) from e
         else:

mlprodict/onnxrt/validate/validate_helper.py

@@ -320,7 +320,6 @@ def default_time_kwargs():
         100: dict(number=8, repeat=10),
         1000: dict(number=5, repeat=5),
         10000: dict(number=3, repeat=3),
-        100000: dict(number=2, repeat=1),
     }