You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
PaddleSpeech/paddlespeech/server/utils/onnx_infer.py

49 lines
1.9 KiB

# Copyright (c) 2021 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
from typing import Optional
import onnxruntime as ort
from .log import logger
def get_sess(model_path: Optional[os.PathLike]=None, sess_conf: dict=None):
logger.info(f"ort sessconf: {sess_conf}")
sess_options = ort.SessionOptions()
sess_options.graph_optimization_level = ort.GraphOptimizationLevel.ORT_ENABLE_ALL
if sess_conf.get('graph_optimization_level', 99) == 0:
sess_options.graph_optimization_level = ort.GraphOptimizationLevel.ORT_DISABLE_ALL
sess_options.execution_mode = ort.ExecutionMode.ORT_SEQUENTIAL
# "gpu:0"
providers = ['CPUExecutionProvider']
if "gpu" in sess_conf.get("device", ""):
providers = ['CUDAExecutionProvider']
# fastspeech2/mb_melgan can't use trt now!
if sess_conf.get("use_trt", 0):
providers = ['TensorrtExecutionProvider']
logger.info(f"ort providers: {providers}")
if 'cpu_threads' in sess_conf:
sess_options.intra_op_num_threads = sess_conf.get("cpu_threads", 0)
else:
sess_options.intra_op_num_threads = sess_conf.get("intra_op_num_threads", 0)
sess_options.inter_op_num_threads = sess_conf.get("inter_op_num_threads", 0)
sess = ort.InferenceSession(
model_path, providers=providers, sess_options=sess_options)
return sess