wanghaoshuang
7 years ago
commit
5993155d67
@ -1,15 +1,106 @@
|
|||||||
// Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
|
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
|
||||||
//
|
|
||||||
// Licensed under the Apache License, Version 2.0 (the "License");
|
Licensed under the Apache License, Version 2.0 (the "License");
|
||||||
// you may not use this file except in compliance with the License.
|
you may not use this file except in compliance with the License.
|
||||||
// You may obtain a copy of the License at
|
You may obtain a copy of the License at
|
||||||
//
|
|
||||||
// http://www.apache.org/licenses/LICENSE-2.0
|
http://www.apache.org/licenses/LICENSE-2.0
|
||||||
//
|
|
||||||
// Unless required by applicable law or agreed to in writing, software
|
Unless required by applicable law or agreed to in writing, software
|
||||||
// distributed under the License is distributed on an "AS IS" BASIS,
|
distributed under the License is distributed on an "AS IS" BASIS,
|
||||||
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||||
// See the License for the specific language governing permissions and
|
See the License for the specific language governing permissions and
|
||||||
// limitations under the License.
|
limitations under the License. */
|
||||||
|
|
||||||
data_type_transform.cc
|
#include "paddle/fluid/framework/data_type_transform.h"
|
||||||
|
|
||||||
|
#include "paddle/fluid/framework/selected_rows.h"
|
||||||
|
#include "paddle/fluid/platform/transform.h"
|
||||||
|
|
||||||
|
namespace paddle {
|
||||||
|
namespace framework {
|
||||||
|
|
||||||
|
template <typename InType, typename OutType>
|
||||||
|
struct CastDataTypeFunctor {
|
||||||
|
HOSTDEVICE inline OutType operator()(InType in) const {
|
||||||
|
return static_cast<OutType>(in);
|
||||||
|
}
|
||||||
|
};
|
||||||
|
|
||||||
|
template <typename InType>
|
||||||
|
struct CastDataType {
|
||||||
|
CastDataType(const framework::Tensor& in, framework::Tensor* out,
|
||||||
|
const platform::DeviceContext* ctx)
|
||||||
|
: in_(in), out_(out), ctx_(ctx) {}
|
||||||
|
const framework::Tensor in_;
|
||||||
|
framework::Tensor* out_;
|
||||||
|
const platform::DeviceContext* ctx_;
|
||||||
|
|
||||||
|
template <typename OutType>
|
||||||
|
void apply() {
|
||||||
|
auto* in_begin = in_.data<InType>();
|
||||||
|
auto* in_end = in_begin + in_.numel();
|
||||||
|
auto* out_begin = out_->mutable_data<OutType>(in_.place());
|
||||||
|
|
||||||
|
if (platform::is_cpu_place(in_.place())) {
|
||||||
|
platform::Transform<platform::CPUDeviceContext> trans;
|
||||||
|
auto* context = static_cast<const platform::CPUDeviceContext*>(ctx_);
|
||||||
|
trans(*context, in_begin, in_end, out_begin,
|
||||||
|
CastDataTypeFunctor<InType, OutType>());
|
||||||
|
#ifdef __NVCC__
|
||||||
|
} else if (platform::is_gpu_place(in_.place())) {
|
||||||
|
platform::Transform<platform::CUDADeviceContext> trans;
|
||||||
|
auto* context = static_cast<const platform::CUDADeviceContext*>(ctx_);
|
||||||
|
trans(*context, in_begin, in_end, out_begin,
|
||||||
|
CastDataTypeFunctor<InType, OutType>());
|
||||||
|
context->Wait();
|
||||||
|
#endif
|
||||||
|
} else {
|
||||||
|
PADDLE_THROW("Unsupported place!");
|
||||||
|
}
|
||||||
|
}
|
||||||
|
};
|
||||||
|
|
||||||
|
void TransDataType(const OpKernelType& kernel_type_for_var,
|
||||||
|
const OpKernelType& expected_kernel_type, const Tensor& in,
|
||||||
|
Tensor* out) {
|
||||||
|
platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance();
|
||||||
|
|
||||||
|
out->Resize(in.dims());
|
||||||
|
auto src_type = kernel_type_for_var.data_type_;
|
||||||
|
auto dst_type = expected_kernel_type.data_type_;
|
||||||
|
auto ctx = pool.Get(in.place());
|
||||||
|
|
||||||
|
switch (src_type) {
|
||||||
|
case proto::VarType::FP16:
|
||||||
|
framework::VisitDataType(dst_type,
|
||||||
|
CastDataType<platform::float16>(in, out, ctx));
|
||||||
|
break;
|
||||||
|
case proto::VarType::FP32:
|
||||||
|
framework::VisitDataType(dst_type, CastDataType<float>(in, out, ctx));
|
||||||
|
break;
|
||||||
|
case proto::VarType::FP64:
|
||||||
|
framework::VisitDataType(dst_type, CastDataType<double>(in, out, ctx));
|
||||||
|
break;
|
||||||
|
case proto::VarType::INT32:
|
||||||
|
framework::VisitDataType(dst_type, CastDataType<int>(in, out, ctx));
|
||||||
|
break;
|
||||||
|
case proto::VarType::INT64:
|
||||||
|
framework::VisitDataType(dst_type, CastDataType<int64_t>(in, out, ctx));
|
||||||
|
break;
|
||||||
|
case proto::VarType::BOOL:
|
||||||
|
framework::VisitDataType(dst_type, CastDataType<bool>(in, out, ctx));
|
||||||
|
break;
|
||||||
|
case proto::VarType::INT16:
|
||||||
|
framework::VisitDataType(dst_type, CastDataType<bool>(in, out, ctx));
|
||||||
|
break;
|
||||||
|
case proto::VarType::UINT8:
|
||||||
|
framework::VisitDataType(dst_type, CastDataType<bool>(in, out, ctx));
|
||||||
|
break;
|
||||||
|
default:
|
||||||
|
PADDLE_THROW("Not support type %d", src_type);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
} // namespace framework
|
||||||
|
} // namespace paddle
|
||||||
|
|||||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
@ -1,97 +0,0 @@
|
|||||||
// Copyright (c) 2018 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.
|
|
||||||
|
|
||||||
#include <chrono>
|
|
||||||
#include <iostream>
|
|
||||||
#include <fstream>
|
|
||||||
#include "paddle/fluid/inference/api/paddle_inference_api.h"
|
|
||||||
|
|
||||||
namespace paddle {
|
|
||||||
|
|
||||||
std::string DIRNAME = "./LB_icnet_model";
|
|
||||||
//std::string DIRNAME = "./infer_models";
|
|
||||||
NativeConfig GetConfig() {
|
|
||||||
NativeConfig config;
|
|
||||||
config.prog_file=DIRNAME + "/__model__";
|
|
||||||
config.param_file=DIRNAME + "/__params__";
|
|
||||||
config.fraction_of_gpu_memory = 0.8;
|
|
||||||
config.use_gpu = true;
|
|
||||||
config.device = 0;
|
|
||||||
return config;
|
|
||||||
}
|
|
||||||
|
|
||||||
using Time = decltype(std::chrono::high_resolution_clock::now());
|
|
||||||
Time time() { return std::chrono::high_resolution_clock::now(); };
|
|
||||||
double time_diff(Time t1, Time t2) {
|
|
||||||
typedef std::chrono::microseconds ms;
|
|
||||||
auto diff = t2 - t1;
|
|
||||||
ms counter = std::chrono::duration_cast<ms>(diff);
|
|
||||||
return counter.count() / 1000.0;
|
|
||||||
}
|
|
||||||
|
|
||||||
void test_naive(int batch_size){
|
|
||||||
NativeConfig config = GetConfig();
|
|
||||||
// config.model_dir = model_path;
|
|
||||||
auto predictor = CreatePaddlePredictor<NativeConfig>(config);
|
|
||||||
int height = 449;
|
|
||||||
int width = 581;
|
|
||||||
//int height = 3;
|
|
||||||
//int width = 3;
|
|
||||||
int num_sum = height * width * 3 * batch_size;
|
|
||||||
|
|
||||||
std::vector<float> data;
|
|
||||||
|
|
||||||
for(int i = 0; i < num_sum; i++) {
|
|
||||||
data.push_back(0.0);
|
|
||||||
}
|
|
||||||
|
|
||||||
PaddleTensor tensor;
|
|
||||||
tensor.shape = std::vector<int>({batch_size, 3, height, width});
|
|
||||||
tensor.data.Resize(sizeof(float) * batch_size * 3 * height * width);
|
|
||||||
std::copy(data.begin(), data.end(), static_cast<float*>(tensor.data.data()));
|
|
||||||
tensor.dtype = PaddleDType::FLOAT32;
|
|
||||||
std::vector<PaddleTensor> paddle_tensor_feeds(1, tensor);
|
|
||||||
PaddleTensor tensor_out;
|
|
||||||
|
|
||||||
std::vector<PaddleTensor> outputs(1, tensor_out);
|
|
||||||
|
|
||||||
predictor->Run(paddle_tensor_feeds, &outputs, batch_size);
|
|
||||||
std::cout << "start predict123:" << std::endl;
|
|
||||||
auto time1 = time();
|
|
||||||
|
|
||||||
for(size_t i = 0; i < 2; i++) {
|
|
||||||
predictor->Run(paddle_tensor_feeds, &outputs, batch_size);
|
|
||||||
std::cout << "pass " << i;
|
|
||||||
}
|
|
||||||
|
|
||||||
auto time2 = time();
|
|
||||||
std::ofstream ofresult("naive_test_result.txt", std::ios::app);
|
|
||||||
|
|
||||||
std::cout <<"batch: " << batch_size << " predict cost: " << time_diff(time1, time2) / 100.0 << "ms" << std::endl;
|
|
||||||
std::cout << outputs.size() << std::endl;
|
|
||||||
/*
|
|
||||||
int64_t * data_o = static_cast<int64_t*>(outputs[0].data.data());
|
|
||||||
for (size_t j = 0; j < outputs[0].data.length() / sizeof(int64_t); ++j) {
|
|
||||||
ofresult << std::to_string(data_o[j]) << " ";
|
|
||||||
}
|
|
||||||
ofresult << std::endl;
|
|
||||||
ofresult.close();
|
|
||||||
*/
|
|
||||||
}
|
|
||||||
} // namespace paddle
|
|
||||||
|
|
||||||
int main(int argc, char** argv) {
|
|
||||||
paddle::test_naive(1 << 0);
|
|
||||||
return 0;
|
|
||||||
}
|
|
||||||
Loading…
Reference in new issue