Add test for Python BLS model loading API (#5980)

* Add test for Python BLS model loading API

* Fix up

qa/L0_backend_python/bls/test.sh

@@ -227,6 +227,98 @@ else
     RET=1
 fi
 
+# Test model loading API with BLS
+rm -fr ./models
+mkdir -p models/bls_model_loading/1/
+cp ../../python_models/bls_model_loading/model.py models/bls_model_loading/1/
+cp ../../python_models/bls_model_loading/config.pbtxt models/bls_model_loading/
+cp -fr ${DATADIR}/qa_model_repository/onnx_int32_int32_int32 models/.
+# Make only version 2, 3 is valid version directory
+rm -rf models/onnx_int32_int32_int32/1
+
+SERVER_LOG="./bls_model_loading_server.log"
+SERVER_ARGS="--model-repository=`pwd`/models --backend-directory=${BACKEND_DIR} --model-control-mode=explicit --log-verbose=1"
+
+run_server
+if [ "$SERVER_PID" == "0" ]; then
+    echo -e "\n***\n*** Failed to start $SERVER\n***"
+    cat $SERVER_LOG
+    exit 1
+fi
+
+export MODEL_NAME='bls_model_loading'
+
+set +e
+code=`curl -s -w %{http_code} -X POST localhost:8000/v2/repository/models/${MODEL_NAME}/load`
+set -e
+if [ "$code" == "400" ]; then
+    echo -e "\n***\n*** Failed to load model '${MODEL_NAME}'\n***"
+    RET=1
+fi
+
+set +e
+
+python3 $CLIENT_PY >> $CLIENT_LOG 2>&1
+if [ $? -ne 0 ]; then
+    echo -e "\n***\n*** 'bls_model_loading' test FAILED. \n***"
+    cat $CLIENT_LOG
+    RET=1
+else
+    check_test_results $TEST_RESULT_FILE $EXPECTED_NUM_TESTS
+    if [ $? -ne 0 ]; then
+        cat $CLIENT_LOG
+        echo -e "\n***\n*** Test Result Verification Failed\n***"
+        RET=1
+    fi
+fi
+
+set -e
+
+kill $SERVER_PID
+wait $SERVER_PID
+
+# Test model loading API with BLS warmup
+(cd models/bls_model_loading && \
+        echo "model_warmup [{" >> config.pbtxt && \
+        echo "    name : \"regular sample\"" >> config.pbtxt && \
+        echo "    batch_size: 1" >> config.pbtxt && \
+        echo "    inputs {" >> config.pbtxt && \
+        echo "        key: \"INPUT0\"" >> config.pbtxt && \
+        echo "        value: {" >> config.pbtxt && \
+        echo "            data_type: TYPE_FP32" >> config.pbtxt && \
+        echo "            dims: 4" >> config.pbtxt && \
+        echo "            zero_data: false" >> config.pbtxt && \
+        echo "        }" >> config.pbtxt && \
+        echo "    }" >> config.pbtxt && \
+        echo "    inputs {" >> config.pbtxt && \
+        echo "        key: \"INPUT1\"" >> config.pbtxt && \
+        echo "        value: {" >> config.pbtxt && \
+        echo "            data_type: TYPE_FP32" >> config.pbtxt && \
+        echo "            dims: 4" >> config.pbtxt && \
+        echo "            zero_data: false" >> config.pbtxt && \
+        echo "        }" >> config.pbtxt && \
+        echo "    }" >> config.pbtxt && \
+        echo "}]" >> config.pbtxt )
+
+SERVER_LOG="./bls_model_loading_server_warmup.log"
+run_server
+if [ "$SERVER_PID" == "0" ]; then
+    echo -e "\n***\n*** Failed to start $SERVER\n***"
+    cat $SERVER_LOG
+    exit 1
+fi
+
+set +e
+code=`curl -s -w %{http_code} -X POST localhost:8000/v2/repository/models/${MODEL_NAME}/load`
+set -e
+if [ "$code" == "400" ]; then
+    echo -e "\n***\n*** Failed to load model '${MODEL_NAME}'\n***"
+    RET=1
+fi
+
+kill $SERVER_PID
+wait $SERVER_PID
+
 if [ $RET -eq 1 ]; then
     cat $CLIENT_LOG
     cat $SERVER_LOG

qa/python_models/bls_model_loading/config.pbtxt

@@ -0,0 +1,43 @@
+# Copyright 2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+name: "bls_model_loading"
+backend: "python"
+
+output [
+  {
+    name: "OUTPUT0"
+    data_type: TYPE_BOOL
+    dims: [ 1 ]
+  }
+]
+
+instance_group [
+  { 
+    count: 3
+    kind: KIND_CPU 
+  }
+]

qa/python_models/bls_model_loading/model.py

@@ -0,0 +1,119 @@
+# Copyright 2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+import numpy as np
+import unittest
+import triton_python_backend_utils as pb_utils
+
+
+class PBBLSModelLoadingTest(unittest.TestCase):
+
+    def setUp(self):
+        self.model_name = "onnx_int32_int32_int32"
+
+    def tearDown(self):
+        pb_utils.unload_model(self.model_name)
+
+    def test_load_unload_model(self):
+        self.assertFalse(pb_utils.is_model_ready(model_name=self.model_name))
+        pb_utils.load_model(model_name=self.model_name)
+        self.assertTrue(pb_utils.is_model_ready(self.model_name))
+        pb_utils.unload_model(self.model_name)
+        self.assertFalse(pb_utils.is_model_ready(self.model_name))
+
+    def test_load_with_config_override(self):
+        self.assertFalse(pb_utils.is_model_ready(self.model_name))
+        pb_utils.load_model(self.model_name)
+        self.assertTrue(pb_utils.is_model_ready(self.model_name))
+
+        # Send the config with the wrong format
+        wrong_config = "\"parameters\": {\"config\": {{\"backend\":\"onnxruntime\", \"version_policy\":{\"specific\":{\"versions\":[2]}}}}}"
+        with self.assertRaises(pb_utils.TritonModelException):
+            pb_utils.load_model(model_name=self.model_name, config=wrong_config)
+        # The model should not be changed after a failed load model request
+        for version in ["2", "3"]:
+            self.assertTrue(pb_utils.is_model_ready(model_name=self.model_name, model_version=version))
+
+        # Send the config with the correct format
+        config = "{\"backend\":\"onnxruntime\", \"version_policy\":{\"specific\":{\"versions\":[2]}}}"
+        pb_utils.load_model(self.model_name, config=config)
+        # The model should be changed after a successful load model request
+        self.assertTrue(pb_utils.is_model_ready(self.model_name, "2"))
+        self.assertFalse(pb_utils.is_model_ready(self.model_name, "3"))
+
+    def test_load_with_file_override(self):
+        self.assertFalse(pb_utils.is_model_ready(self.model_name))
+        pb_utils.load_model(self.model_name)
+        self.assertTrue(pb_utils.is_model_ready(self.model_name))
+
+        override_name = "override_model"
+        config = "{\"backend\":\"onnxruntime\"}"
+        with open('models/onnx_int32_int32_int32/3/model.onnx', 'rb') as file:
+            data = file.read()
+        files = {"file:1/model.onnx": data}
+
+        # Request to load the model with override file, should fail without
+        # providing override config.
+        with self.assertRaises(pb_utils.TritonModelException):
+            pb_utils.load_model(self.model_name, "", files)
+
+        # Request to load the model with override file and config in a different name
+        pb_utils.load_model(model_name=override_name, config=config, files=files)
+        # Sanity check that the model with original name is unchanged
+        self.assertFalse(pb_utils.is_model_ready(self.model_name, "1"))
+        self.assertTrue(pb_utils.is_model_ready(self.model_name, "3"))
+
+        # Check the override model readiness
+        self.assertTrue(pb_utils.is_model_ready(override_name, "1"))
+        self.assertFalse(pb_utils.is_model_ready(override_name, "3"))
+
+        # Request to load the model with override file and config in original name
+        pb_utils.load_model(self.model_name, config, files)
+        # Check that the model with original name is changed
+        self.assertTrue(pb_utils.is_model_ready(self.model_name, "1"))
+        self.assertFalse(pb_utils.is_model_ready(self.model_name, "3"))
+
+        # Sanity check readiness of the different named model
+        self.assertTrue(pb_utils.is_model_ready(override_name, "1"))
+        self.assertFalse(pb_utils.is_model_ready(override_name, "3"))
+
+
+class TritonPythonModel:
+
+    def initialize(self, args):
+        # Run the unittest during initialization
+        test = unittest.main('model', exit=False)
+        self.result = test.result.wasSuccessful()
+
+    def execute(self, requests):
+        responses = []
+        for _ in requests:
+            responses.append(
+                pb_utils.InferenceResponse([
+                    pb_utils.Tensor('OUTPUT0',
+                                    np.array([self.result], dtype=np.float16))
+                ]))
+        return responses