Fix Training Packaging pipeline (#9885)

* Fix Training Packaging pipeline

docs/python/inference/examples/plot_common_errors.py

@@ -21,7 +21,7 @@ import numpy
 from onnxruntime.datasets import get_example
 
 example2 = get_example("logreg_iris.onnx")
-sess = rt.InferenceSession(example2)
+sess = rt.InferenceSession(example2, providers=rt.get_available_providers())
 
 input_name = sess.get_inputs()[0].name
 output_name = sess.get_outputs()[0].name

docs/python/inference/examples/plot_convert_pipeline_vectorizer.py

@@ -72,7 +72,7 @@ with open("pipeline_vectorize.onnx", "wb") as f:
 import onnxruntime as rt
 from onnxruntime.capi.onnxruntime_pybind11_state import InvalidArgument
 
-sess = rt.InferenceSession("pipeline_vectorize.onnx")
+sess = rt.InferenceSession("pipeline_vectorize.onnx", providers=rt.get_available_providers())
 
 import numpy
 inp, out = sess.get_inputs()[0], sess.get_outputs()[0]

docs/python/inference/examples/plot_load_and_predict.py

@@ -21,7 +21,7 @@ from onnxruntime.datasets import get_example
 # The model is available on github `onnx...test_sigmoid <https://github.com/onnx/onnx/tree/master/onnx/backend/test/data/node/test_sigmoid>`_.
 
 example1 = get_example("sigmoid.onnx")
-sess = rt.InferenceSession(example1)
+sess = rt.InferenceSession(example1, providers=rt.get_available_providers())
 
 #########################
 # Let's see the input name and shape.

docs/python/inference/examples/plot_metadata.py

@@ -31,8 +31,8 @@ print("producer_version={}".format(model.producer_version))
 #############################
 # With *ONNX Runtime*:
 
-from onnxruntime import InferenceSession
-sess = InferenceSession(example)
+import onnxruntime as rt
+sess = rt.InferenceSession(example, providers=rt.get_available_providers())
 meta = sess.get_modelmeta()
 
 print("custom_metadata_map={}".format(meta.custom_metadata_map))

docs/python/inference/examples/plot_profiling.py

@@ -35,7 +35,7 @@ def change_ir_version(filename, ir_version=6):
 example1 = get_example("mul_1.onnx")
 onnx_model = change_ir_version(example1)
 onnx_model_str = onnx_model.SerializeToString()
-sess = rt.InferenceSession(onnx_model_str)
+sess = rt.InferenceSession(onnx_model_str, providers=rt.get_available_providers())
 input_name = sess.get_inputs()[0].name
 
 x = numpy.array([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]], dtype=numpy.float32)
@@ -48,7 +48,7 @@ print(res)
 
 options = rt.SessionOptions()
 options.enable_profiling = True
-sess_profile = rt.InferenceSession(onnx_model_str, options)
+sess_profile = rt.InferenceSession(onnx_model_str, options, providers=rt.get_available_providers())
 input_name = sess.get_inputs()[0].name
 
 x = numpy.array([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]], dtype=numpy.float32)

docs/python/inference/examples/plot_train_convert_predict.py

@@ -64,7 +64,7 @@ with open("logreg_iris.onnx", "wb") as f:
 # its input and output.
 
 import onnxruntime as rt
-sess = rt.InferenceSession("logreg_iris.onnx")
+sess = rt.InferenceSession("logreg_iris.onnx", providers=rt.get_available_providers())
 
 print("input name='{}' and shape={}".format(
     sess.get_inputs()[0].name, sess.get_inputs()[0].shape))
@@ -180,7 +180,7 @@ with open("rf_iris.onnx", "wb") as f:
 ###################################
 # We compare.
 
-sess = rt.InferenceSession("rf_iris.onnx")
+sess = rt.InferenceSession("rf_iris.onnx", providers=rt.get_available_providers())
 
 def sess_predict_proba_rf(x):
     return sess.run([prob_name], {input_name: x.astype(numpy.float32)})[0]
@@ -204,7 +204,7 @@ for n_trees in range(5, 51, 5):
     onx = convert_sklearn(rf, initial_types=initial_type)
     with open("rf_iris_%d.onnx" % n_trees, "wb") as f:
         f.write(onx.SerializeToString())
-    sess = rt.InferenceSession("rf_iris_%d.onnx" % n_trees)
+    sess = rt.InferenceSession("rf_iris_%d.onnx" % n_trees, providers=rt.get_available_providers())
     def sess_predict_proba_loop(x):
         return sess.run([prob_name], {input_name: x.astype(numpy.float32)})[0]
     tsk = speed("loop(X_test, rf.predict_proba, 100)", number=5, repeat=5)

docs/python/inference/tutorial.rst

@@ -82,7 +82,7 @@ for this machine learning model.
     import numpy
     import onnxruntime as rt
 
-    sess = rt.InferenceSession("logreg_iris.onnx")
+    sess = rt.InferenceSession("logreg_iris.onnx", providers=rt.get_available_providers())
     input_name = sess.get_inputs()[0].name
     pred_onx = sess.run(None, {input_name: X_test.astype(numpy.float32)})[0]
     print(pred_onx)
@@ -97,7 +97,7 @@ by specifying its name into a list.
     import numpy
     import onnxruntime as rt
 
-    sess = rt.InferenceSession("logreg_iris.onnx")
+    sess = rt.InferenceSession("logreg_iris.onnx", providers=rt.get_available_providers())
     input_name = sess.get_inputs()[0].name
     label_name = sess.get_outputs()[0].name
     pred_onx = sess.run([label_name], {input_name: X_test.astype(numpy.float32)})[0]

onnxruntime/test/python/onnxruntime_test_python_keras.py

@@ -68,7 +68,7 @@ class TestInferenceSessionKeras(unittest.TestCase):
 
         # runtime
         content = converted_model.SerializeToString()
-        rt = onnxrt.InferenceSession(content)
+        rt = onnxrt.InferenceSession(content, providers=onnxrt.get_available_providers())
         input = {rt.get_inputs()[0].name: x}
         actual_rt = rt.run(None, input)
         self.assertEqual(len(actual_rt), 1)

orttraining/tools/scripts/layer_norm_transform.py

@@ -163,11 +163,11 @@ def main():
     input_mask = np.ones((batch, sq_length), dtype=np.int64)
 
     # Do forward using the original model.
-    sess = ort.InferenceSession(model_file_path)
+    sess = ort.InferenceSession(model_file_path, providers=ort.get_available_providers())
     result = sess.run(None, {'input1': input_ids, 'input2': segment_ids, 'input3': input_mask})
 
     # Do forward using the new model.
-    new_sess = ort.InferenceSession(new_model_file_path)
+    new_sess = ort.InferenceSession(new_model_file_path, providers=ort.get_available_providers())
     new_result = new_sess.run(None, {'input1': input_ids, 'input2': segment_ids, 'input3': input_mask})
 
     # Compare the outcomes from the two models.

orttraining/tools/scripts/model_transform.py

@@ -298,11 +298,11 @@ segment_ids = np.random.randint(low=0, high=2, size=(batch, sq_length), dtype=np
 input_mask = np.ones((batch, sq_length), dtype=np.int64)
 
 # Do forward using the original model.
-sess = ort.InferenceSession(input_model_name)
+sess = ort.InferenceSession(input_model_name, providers=ort.get_available_providers())
 result = sess.run(None, {'input1': input_ids, 'input2': segment_ids, 'input3': input_mask})
 
 # Do forward using the new model.
-new_sess = ort.InferenceSession(output_model_name)
+new_sess = ort.InferenceSession(output_model_name, providers=ort.get_available_providers())
 new_result = new_sess.run(None, {'input1': input_ids, 'input2': segment_ids, 'input3': input_mask})
 
 # Compare the outcomes from the two models.

orttraining/tools/scripts/nv_run_pretraining.py

@@ -528,7 +528,7 @@ def main():
                       is_model_exported = False
 
                       import onnxruntime as ort
-                      sess = ort.InferenceSession(onnx_path)
+                      sess = ort.InferenceSession(onnx_path, providers=ort.get_available_providers())
                       result = sess.run(None, {'input1': input_ids.cpu().numpy(), 'input2': segment_ids.cpu().numpy(), 'input3': input_mask.cpu().numpy()})
 
                       print('---ORT result---')