wehub-resource-sync
145 lines
5.0 KiB
Python
Executable File
145 lines
5.0 KiB
Python
Executable File
# Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
import os
|
|
import unittest
|
|
|
|
import numpy as np
|
|
|
|
import paddle
|
|
from paddle.base.core import AnalysisConfig, create_paddle_predictor
|
|
from paddle.jit import to_static
|
|
|
|
|
|
class OpenVINOBaseTest(unittest.TestCase):
|
|
def __init__(self, methodName='runTest'):
|
|
super().__init__(methodName)
|
|
paddle.device.set_device("cpu")
|
|
self.batch_size = 1
|
|
self.infer_threads = 1
|
|
self.precision = "float32"
|
|
current_file_path = os.path.abspath(__file__)
|
|
current_dir = os.path.dirname(current_file_path)
|
|
self.temp_dir = current_dir
|
|
self.model_name = "model"
|
|
self.model_dir = None
|
|
self.paddle_config = None
|
|
self.openvino_config = None
|
|
self.input_names = None
|
|
self.input_shape_map = None
|
|
self.output_names = None
|
|
self.input_data = []
|
|
self.output_expected = []
|
|
self.output_openvino = []
|
|
self.precision_map = {
|
|
"int8": AnalysisConfig.Int8,
|
|
"float16": AnalysisConfig.Half,
|
|
"float32": AnalysisConfig.Float32,
|
|
}
|
|
|
|
def to_static(self, model, input_spec):
|
|
self.model_dir = os.path.join(self.temp_dir, self.model_name)
|
|
net = to_static(
|
|
model,
|
|
input_spec=input_spec,
|
|
full_graph=True,
|
|
)
|
|
paddle.jit.save(net, os.path.join(self.model_dir, 'inference'))
|
|
|
|
def prepare_paddle_config(self):
|
|
if self.paddle_config is not None:
|
|
return
|
|
self.paddle_config = AnalysisConfig(
|
|
os.path.join(self.model_dir, 'inference.pdmodel'),
|
|
os.path.join(self.model_dir, 'inference.pdiparams'),
|
|
)
|
|
self.paddle_config.disable_gpu()
|
|
self.paddle_config.switch_ir_optim(False)
|
|
|
|
def prepare_openvino_config(self):
|
|
if self.openvino_config is not None:
|
|
return
|
|
self.openvino_config = AnalysisConfig(
|
|
os.path.join(self.model_dir, 'inference.pdmodel'),
|
|
os.path.join(self.model_dir, 'inference.pdiparams'),
|
|
)
|
|
self.openvino_config.disable_gpu()
|
|
self.openvino_config.enable_openvino_engine(
|
|
self.precision_map[self.precision]
|
|
)
|
|
self.openvino_config.set_cpu_math_library_num_threads(
|
|
self.infer_threads
|
|
)
|
|
cache_dir = os.path.join(self.model_dir, '__cache__')
|
|
self.openvino_config.set_optim_cache_dir(cache_dir)
|
|
|
|
def prepare_input(self):
|
|
if len(self.input_data) != len(self.input_names):
|
|
for name in self.input_names:
|
|
new_shape = [
|
|
self.batch_size if x == -1 else x
|
|
for x in self.input_shape_map[name]
|
|
]
|
|
self.input_data.append(
|
|
np.random.random(new_shape).astype("float32")
|
|
)
|
|
|
|
def run_paddle(self):
|
|
if self.paddle_config is None:
|
|
self.prepare_paddle_config()
|
|
self.paddle_predictor = create_paddle_predictor(self.paddle_config)
|
|
self.input_names = self.paddle_predictor.get_input_names()
|
|
self.input_shape_map = self.paddle_predictor.get_input_tensor_shape()
|
|
|
|
self.prepare_input()
|
|
|
|
for idx, name in enumerate(self.input_names):
|
|
tensor = self.paddle_predictor.get_input_tensor(name)
|
|
tensor.copy_from_cpu(self.input_data[idx])
|
|
|
|
self.paddle_predictor.zero_copy_run()
|
|
|
|
self.output_names = self.paddle_predictor.get_output_names()
|
|
for name in self.output_names:
|
|
self.output_expected.append(
|
|
self.paddle_predictor.get_output_tensor(name).copy_to_cpu()
|
|
)
|
|
|
|
def run_openvino(self):
|
|
self.prepare_openvino_config()
|
|
self.openvino_predictor = create_paddle_predictor(self.openvino_config)
|
|
|
|
for idx, name in enumerate(self.input_names):
|
|
tensor = self.openvino_predictor.get_input_tensor(name)
|
|
tensor.copy_from_cpu(self.input_data[idx])
|
|
|
|
self.openvino_predictor.zero_copy_run()
|
|
|
|
for name in self.output_names:
|
|
self.output_openvino.append(
|
|
self.paddle_predictor.get_output_tensor(name).copy_to_cpu()
|
|
)
|
|
|
|
def check_result(self, rtol=1e-3, atol=1e-3):
|
|
self.run_paddle()
|
|
self.run_openvino()
|
|
|
|
for i in range(len(self.output_expected)):
|
|
np.testing.assert_allclose(
|
|
self.output_expected[i],
|
|
self.output_openvino[i],
|
|
rtol=rtol,
|
|
atol=atol,
|
|
)
|