def _main():
base_dir = os.path.dirname(__file__)
request = {
"serving_type": "openvino",
"input_model": os.path.join(base_dir, "./model/mnist.onnx"),
"export_path": os.path.join(base_dir, "model_repos"),
"input_names": [
"input.1"
],
"input_formats": [
"channels_first"
],
"output_names": [
"20"
],
"input_signatures": [
"image"
],
"output_signatures": [
"label"
],
"model_name": "mnist",
"job_id": "mnist_pytorch",
"callback": "",
"max_batch_size": 128
}
result = model_compiler.compile_model(request)
print(result)
def test_success_result(self):
for k in [keras, tf.keras]:
with self.subTest(k=k), NamedTemporaryFile(
suffix='.h5') as model_file, TemporaryDirectory(
) as target_dir:
with tf.compat.v1.Session(graph=tf.Graph()):
k.Sequential([k.layers.Dense(units=4, input_shape=[8])
]).save(model_file.name)
result = model_compiler.compile_model({
'serving_type':
'tf',
'model_name':
'foobar',
'version':
4,
'max_batch_size':
7,
'input_model':
model_file.name,
'input_signatures': ['x'],
'output_signatures': ['y'],
'export_path':
target_dir
})
self.assertEqual(
result, {
'status': 'success',
'path': path.join(target_dir, 'foobar_4.zip')
})
def test_error_result(self):
result = model_compiler.compile_model({})
self.assertEqual(
result, {
'status': 'failure',
'error_msg': 'Unable to determine the source model type.'
})
def compile_model(args):
"""
Compile the model.
"""
request_dir = os.path.join(args.test_model_path, args.serving_model_json)
try:
with open(request_dir, 'r') as request_file:
test_model_dir = args.test_model_path
request = _get_request(request_file, test_model_dir)
result = model_compiler.compile_model(request)
print(result)
except FileNotFoundError:
print(f"Can not compile the model in {os.path.join(test_model_dir, args.model_path)}")
def _main():
base_dir = os.path.dirname(__file__)
request = {
"serving_type": "tflite", # or openvino, tensorrt
"input_model": os.path.join(base_dir, "./model/resnet50.h5"),
"export_path": os.path.join(base_dir, "model_repos"),
"input_formats": ["channels_last"],
"input_signatures": ["image"],
"output_signatures": ["label"],
"optimization": True,
"model_name": "resnet50",
"max_batch_size": 1
}
result = model_compiler.compile_model(request)
print(result)
def _main():
base_dir = os.path.dirname(__file__)
request = {
"serving_type": "tf", # or openvino, tensorrt
"input_model": os.path.join(base_dir, "./model/mnist.h5"),
"export_path": os.path.join(base_dir, "model_repos"),
"input_layer_names": ["conv2d_1"],
"output_layer_names": ["dense_2"],
"input_formats": ["channels_last"],
"input_signatures": ["image"],
"output_signatures": ["label"],
"model_name": "mnist",
"max_batch_size": 128
}
result = model_compiler.compile_model(request)
print(result)
def _main():
base_dir = os.path.dirname(__file__)
request = {
"serving_type": "openvino",
"input_model": os.path.join(base_dir, "./model/resnet50.onnx"),
"export_path": os.path.join(base_dir, "model_repos"),
"input_names": ["image"],
"input_formats": ["channels_first"],
"output_names": ["save_infer_model/scale_0.tmp_0"],
"input_signatures": ["image"],
"output_signatures": ["label"],
"model_name": "ResNet50_vd_ssld",
"job_id": "ResNet50_PaddlePaddle",
"callback": "",
"max_batch_size": 128
}
result = model_compiler.compile_model(request)
print(result)