apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-06 17:17:14 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Fix bug of get input/output information from PaddleBackend (#339)

Browse Source 
* Fix bug of get input/output information from PaddleBackend

* Support Paddle Inference with TensorRT (#340)

* Fix bug
This commit is contained in:
Jason
2022-10-12 11:37:26 +08:00
committed by GitHub
parent 3faaeaea36
commit 945e197bd1
8 changed files with 152 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
fastdeploy/backends/paddle/paddle_backend.h
View File

@@ -20,9 +20,15 @@
#include <vector>
#include "fastdeploy/backends/backend.h"
#ifdef ENABLE_PADDLE_FRONTEND
#include "paddle2onnx/converter.h"
#endif
#include "paddle_inference_api.h" // NOLINT
#ifdef ENABLE_TRT_BACKEND
#include "fastdeploy/backends/tensorrt/trt_backend.h"
#endif
namespace fastdeploy {
struct PaddleBackendOption {
@@ -35,6 +41,11 @@ struct PaddleBackendOption {
bool enable_log_info = false;
bool enable_trt = false;
#ifdef ENABLE_TRT_BACKEND
TrtBackendOption trt_option;
#endif
int mkldnn_cache_size = 1;
int cpu_thread_num = 8;
// initialize memory size(MB) for GPU
@@ -58,18 +69,21 @@ void CopyTensorToCpu(std::unique_ptr<paddle_infer::Tensor>& tensor,
// Convert data type from paddle inference to fastdeploy
FDDataType PaddleDataTypeToFD(const paddle_infer::DataType& dtype);
// Convert data type from paddle2onnx::PaddleReader to fastdeploy
FDDataType ReaderDataTypeToFD(int32_t dtype);
class PaddleBackend : public BaseBackend {
public:
PaddleBackend() {}
virtual ~PaddleBackend() = default;
void BuildOption(const PaddleBackendOption& option,
const std::string& model_file);
void BuildOption(const PaddleBackendOption& option);
bool InitFromPaddle(
const std::string& model_file, const std::string& params_file,
const PaddleBackendOption& option = PaddleBackendOption());
bool Infer(std::vector<FDTensor>& inputs, std::vector<FDTensor>* outputs) override;
bool Infer(std::vector<FDTensor>& inputs,
std::vector<FDTensor>* outputs) override;
int NumInputs() const override { return inputs_desc_.size(); }