p85971406
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!46170 [MS]{Lite][Task] solve modelzoo problems 

Merge pull request !46170 from 刘力力/master_develop_merge
This commit is contained in:
i-robot 2022-11-29 08:35:33 +00:00 committed by Gitee
parent 6d9fdaacc1 3ef65295db
commit 81cc1a9f8a
No known key found for this signature in database
GPG Key ID: 173E9B9CA92EEF8F
3 changed files with 41 additions and 47 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

80
mindspore/lite/src/litert/kernel/cpu/base/select.cc
View File

@ -27,6 +27,7 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Select;
namespace mindspore::kernel {
constexpr static int kConditionIdx = 0;
constexpr static int kFirstIdx = 1;
constexpr static int kSecondIdx = 2;
@ -106,6 +107,25 @@ int MoveData(const std::vector<lite::Tensor *>::iterator &dst_begin,
return RET_OK;
}
template <typename T>
int SelectRun(std::vector<lite::Tensor *> in_tensors, std::vector<lite::Tensor *> out_tensors) {
MS_CHECK_TRUE_MSG(in_tensors.at(kFirstIdx)->Size() == out_tensors.at(0)->Size(), RET_ERROR,
"The tensor size should be the same.");
auto size = in_tensors.at(kFirstIdx)->ElementsNum();
MS_CHECK_GT(size, 0, RET_ERROR);
auto condition = static_cast<bool *>(in_tensors.at(kConditionIdx)->data());
auto input1 = static_cast<T *>(in_tensors.at(kFirstIdx)->data());
auto input2 = static_cast<T *>(in_tensors.at(kSecondIdx)->data());
auto output = static_cast<T *>(out_tensors.at(0)->data());
if (condition == nullptr || input1 == nullptr || input2 == nullptr || output == nullptr) {
return RET_NULL_PTR;
}
for (int i = 0; i < size; i++) {
output[i] = condition[i] ? input1[i] : input2[i];
}
return RET_OK;
}
// inputs: bool*1 true-data*n false-data*n
// output: data*n
int SelectCPUKernel::Run() {
@ -138,52 +158,32 @@ int SelectCPUKernel::Run() {
}
}
} else {
MS_CHECK_TRUE_MSG(bool_tensor->shape().size() == in_tensors_.at(1)->shape().size(), RET_ERROR,
MS_CHECK_TRUE_MSG(bool_tensor->shape().size() == in_tensors_.at(kFirstIdx)->shape().size(), RET_ERROR,
"The tensor size should be the same.");
for (size_t i = 0; i < in_tensors_.at(1)->shape().size(); i++) {
if (bool_tensor->shape()[i] != in_tensors_.at(1)->shape()[i]) {
for (size_t i = 0; i < in_tensors_.at(kFirstIdx)->shape().size(); i++) {
if (bool_tensor->shape()[i] != in_tensors_.at(kFirstIdx)->shape()[i]) {
MS_LOG(ERROR) << "Tensor shapes differ in dim: " << i << " in_tensors_.at(0): " << bool_tensor->shape()[i]
<< " in_tensors_.at(1): " << in_tensors_.at(1)->shape()[i];
<< " in_tensors_.at(1): " << in_tensors_.at(kFirstIdx)->shape()[i];
return RET_ERROR;
}
}
MS_CHECK_TRUE_MSG(in_tensors_.at(1)->Size() == out_tensors_.at(0)->Size(), RET_ERROR,
"The tensor size should be the same.");
auto size = in_tensors_.at(1)->ElementsNum();
MS_CHECK_GT(size, 0, RET_ERROR);
auto condition = static_cast<bool *>(bool_tensor->data());
auto input1 = static_cast<float *>(in_tensors_.at(kFirstIdx)->data());
auto input2 = static_cast<float *>(in_tensors_.at(kSecondIdx)->data());
auto output = static_cast<float *>(out_tensors_.at(0)->data());
if (condition == nullptr || input1 == nullptr || input2 == nullptr || output == nullptr) {
return RET_NULL_PTR;
switch (in_tensors_.at(kFirstIdx)->data_type()) {
case kNumberTypeInt32:
if (SelectRun<int>(in_tensors_, out_tensors_) != RET_OK) {
MS_LOG(ERROR) << "Select Run with integer failed";
return RET_ERROR;
}
break;
case kNumberTypeFloat32:
if (SelectRun<float>(in_tensors_, out_tensors_) != RET_OK) {
MS_LOG(ERROR) << "Select Run with float failed";
return RET_ERROR;
}
break;
default:
MS_LOG(ERROR) << "Unsupported data type for select " << in_tensors_.at(kFirstIdx)->data_type();
return RET_ERROR;
}
auto ret = CheckTensor();
if (ret != RET_OK) {
return ret;
}
for (int i = 0; i < size; i++) {
output[i] = condition[i] ? input1[i] : input2[i];
}
}
return RET_OK;
}
int SelectCPUKernel::CheckTensor() {
auto condition = in_tensors_.at(FIRST_INPUT);
if (condition->data_type() != kNumberTypeBool) {
MS_LOG(ERROR) << "wrong data type!";
return RET_ERROR;
}
auto input1_tensor = in_tensors_.at(SECOND_INPUT);
if (input1_tensor->data_type() != kNumberTypeFloat32) {
MS_LOG(ERROR) << "wrong data type!";
return RET_ERROR;
}
auto input2_tensor = in_tensors_.at(THIRD_INPUT);
if (input2_tensor->data_type() != kNumberTypeFloat32) {
MS_LOG(ERROR) << "wrong data type!";
return RET_ERROR;
}
return RET_OK;
}

3
mindspore/lite/src/litert/kernel/cpu/base/select.h
View File

@ -30,9 +30,6 @@ class SelectCPUKernel : public LiteKernel {
int Prepare() override;
int ReSize() override;
int Run() override;
private:
int CheckTensor();
};
} // namespace mindspore::kernel

5
mindspore/lite/src/litert/lite_session.cc
View File

@ -1704,12 +1704,9 @@ int lite::LiteSession::LoadModelAndCompileByBuf(const char *model_buf, mindspore
MS_LOG(ERROR) << "Import model failed";
return RET_ERROR;
}
(reinterpret_cast<lite::LiteModel *>(model))->set_keep_model_buf(keep_model_buf_);
auto ret = CompileGraph(model);
model->buf = nullptr;
// if (buf_model_type == mindspore::ModelType::kMindIR) {
// delete[] lite_buf;
// lite_buf = nullptr;
// }
if (ret != lite::RET_OK) {
MS_LOG(ERROR) << "Compile model failed";
delete model;