apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-16 13:41:30 +08:00
Code Issues Actions 6 Packages Projects Releases Wiki Activity

[Other] Deprecate some option api and parameters (#1243)

Browse Source 
* Optimize Poros backend

* fix error

* Add more pybind

* fix conflicts

* add some deprecate notices

* [Other] Deprecate some apis in RuntimeOption (#1240)

* Deprecate more options

* modify serving

* Update option.h

* fix tensorrt error

* Update option_pybind.cc

* Update option_pybind.cc

* Fix error in serving

* fix word spell error
This commit is contained in:
Jason
2023-02-07 17:57:46 +08:00
committed by GitHub
parent a18cc0f94c
commit 713afe7f1c
15 changed files with 380 additions and 229 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
fastdeploy/runtime/backends/ort/option.h
View File

@@ -22,20 +22,30 @@
#include <map>
namespace fastdeploy {
/*! @brief Option object to configure ONNX Runtime backend
*/
struct OrtBackendOption {
// -1 means default
// 0: ORT_DISABLE_ALL
// 1: ORT_ENABLE_BASIC
// 2: ORT_ENABLE_EXTENDED
// 99: ORT_ENABLE_ALL (enable some custom optimizations e.g bert)
/*
* @brief Level of graph optimization, -1: mean default(Enable all the optimization strategy)/0: disable all the optimization strategy/1: enable basic strategy/2:enable extend strategy/99: enable all
*/
int graph_optimization_level = -1;
/*
* @brief Number of threads to execute the operator, -1: default
*/
int intra_op_num_threads = -1;
/*
* @brief Number of threads to execute the graph, -1: default. This parameter only will bring effects while the `OrtBackendOption::execution_mode` set to 1.
*/
int inter_op_num_threads = -1;
// 0: ORT_SEQUENTIAL
// 1: ORT_PARALLEL
/*
* @brief Execution mode for the graph, -1: default(Sequential mode)/0: Sequential mode, execute the operators in graph one by one. /1: Parallel mode, execute the operators in graph parallelly.
*/
int execution_mode = -1;
/// Inference device, OrtBackend supports CPU/GPU
Device device = Device::CPU;
/// Inference device id
int device_id = 0;
void* external_stream_ = nullptr;
};
} // namespace fastdeploy