apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-06 00:57:33 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Example] Update runtime examples (#1542)

Browse Source 
* Add notes for tensors

* Optimize some apis

* move some warnings
This commit is contained in:
Jason
2023-03-08 16:56:04 +08:00
committed by GitHub
parent 3d31834193
commit 6be2c0367b
17 changed files with 425 additions and 277 deletions
Download Patch File Download Diff File Expand all files Collapse all files

82
examples/runtime/cpp/infer_paddle_tensorrt.cc
View File

@@ -13,49 +13,61 @@
// limitations under the License.
#include "fastdeploy/runtime.h"
#include <cassert>
namespace fd = fastdeploy;
int main(int argc, char* argv[]) {
std::string model_file = "mobilenetv2/inference.pdmodel";
std::string params_file = "mobilenetv2/inference.pdiparams";
// Download from https://bj.bcebos.com/paddle2onnx/model_zoo/pplcnet.tar.gz
std::string model_file = "pplcnet/inference.pdmodel";
std::string params_file = "pplcnet/inference.pdiparams";
// setup option
// configure runtime
// How to configure by RuntimeOption, refer its api doc for more information
// https://baidu-paddle.github.io/fastdeploy-api/cpp/html/structfastdeploy_1_1RuntimeOption.html
fd::RuntimeOption runtime_option;
runtime_option.SetModelPath(model_file, params_file, fd::ModelFormat::PADDLE);
runtime_option.UseGpu(0);
runtime_option.SetModelPath(model_file, params_file);
runtime_option.UseTrtBackend();
runtime_option.EnablePaddleToTrt();
// init runtime
std::unique_ptr<fd::Runtime> runtime =
std::unique_ptr<fd::Runtime>(new fd::Runtime());
if (!runtime->Init(runtime_option)) {
std::cerr << "--- Init FastDeploy Runitme Failed! "
<< "\n--- Model: " << model_file << std::endl;
return -1;
} else {
std::cout << "--- Init FastDeploy Runitme Done! "
<< "\n--- Model: " << model_file << std::endl;
}
// init input tensor shape
fd::TensorInfo info = runtime->GetInputInfo(0);
info.shape = {1, 3, 224, 224};
std::vector<fd::FDTensor> input_tensors(1);
std::vector<fd::FDTensor> output_tensors(1);
std::vector<float> inputs_data;
inputs_data.resize(1 * 3 * 224 * 224);
for (size_t i = 0; i < inputs_data.size(); ++i) {
inputs_data[i] = std::rand() % 1000 / 1000.0f;
}
input_tensors[0].SetExternalData({1, 3, 224, 224}, fd::FDDataType::FP32, inputs_data.data());
//get input name
input_tensors[0].name = info.name;
// Use NVIDIA GPU to inference
// If need to configure TensorRT backend for more option, we can configure runtime_option.trt_option
// refer https://baidu-paddle.github.io/fastdeploy-api/cpp/html/structfastdeploy_1_1TrtBackendOption.html
runtime_option.UseGpu(0);
// Use float16 inference to improve performance
runtime_option.trt_option.enable_fp16 = true;
// Cache trt engine to reduce time cost in model initialize
runtime_option.trt_option.serialize_file = "./pplcnet_model.trt";
runtime->Infer(input_tensors, &output_tensors);
fd::Runtime runtime;
assert(runtime.Init(runtime_option));
output_tensors[0].PrintInfo();
// Get model's inputs information
// API doc refer https://baidu-paddle.github.io/fastdeploy-api/cpp/html/structfastdeploy_1_1Runtime.html
std::vector<fd::TensorInfo> inputs_info = runtime.GetInputInfos();
// Create dummy data fill with 0.5
std::vector<float> dummy_data(1 * 3 * 224 * 224, 0.5);
// Create inputs/outputs tensors
std::vector<fd::FDTensor> inputs(inputs_info.size());
std::vector<fd::FDTensor> outputs;
// Initialize input tensors
// API doc refer https://baidu-paddle.github.io/fastdeploy-api/cpp/html/structfastdeploy_1_1FDTensor.html
inputs[0].SetData({1, 3, 224, 224}, fd::FDDataType::FP32, dummy_data.data());
inputs[0].name = inputs_info[0].name;
// Inference
assert(runtime.Infer(inputs, &outputs));
// Print debug information of outputs
outputs[0].PrintInfo();
// Get data pointer and print it's elements
const float* data_ptr = reinterpret_cast<const float*>(outputs[0].GetData());
for (size_t i = 0; i < 10 && i < outputs[0].Numel(); ++i) {
std::cout << data_ptr[i] << " ";
}
std::cout << std::endl;
return 0;
}
}