Skip to content

[cherrypick][inference][trt]remove trt sparse weights flags (#53562) #53850

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
May 19, 2023
Merged

[cherrypick][inference][trt]remove trt sparse weights flags (#53562) #53850

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e634a09
remove kSPARSE_WEIGHTS
zhangjun May 17, 2023
beb2256
remove kFASTER_DYNAMIC_SHAPES_0805 and add 'TrtMajorVersion' function
zhangjun May 17, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 6 additions & 9 deletions paddle/fluid/inference/analysis/ir_passes/tensorrt_subgraph_pass.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -556,17 +556,14 @@ void TensorRtSubgraphPass::CreateTensorRTOp(
opt_input_shape = {};
}

auto to_major_version = [&](int full_version) -> float {
return (full_version / 100) / 10.0;
};
const float compile_time_trt_version = to_major_version(TRT_VERSION);
const float run_time_trt_version =
to_major_version(tensorrt::GetInferLibVersion());
if (compile_time_trt_version != run_time_trt_version) {
const float trt_compile_version = tensorrt::TrtMajorVersion(TRT_VERSION);
const float trt_runtime_version =
tensorrt::TrtMajorVersion(tensorrt::GetInferLibVersion());
if (trt_compile_version != trt_runtime_version) {
LOG_FIRST_N(WARNING, 1)
<< "The Paddle Inference library is compiled with "
<< compile_time_trt_version << " version TensorRT, "
<< "but the runtime TensorRT you are using is " << run_time_trt_version
<< trt_compile_version << " version TensorRT, "
<< "but the runtime TensorRT you are using is " << trt_runtime_version
<< " version. "
"This might cause serious compatibility issues. We strongly "
"recommend using the same TRT version at runtime.";
Expand Down
7 changes: 0 additions & 7 deletions paddle/fluid/inference/tensorrt/engine.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -158,12 +158,6 @@ void TensorRTEngine::FreezeNetwork() {
infer_builder_config_->setMaxWorkspaceSize(max_workspace_);
#endif

#if IS_TRT_VERSION_GE(8500)
infer_builder_config_->setPreviewFeature(
nvinfer1::PreviewFeature::kFASTER_DYNAMIC_SHAPES_0805, true);
#else
#endif

bool enable_fp16 = (precision_ == AnalysisConfig::Precision::kHalf);
if (enable_fp16) {
bool support_fp16 = infer_builder_->platformHasFastFp16();
Expand Down Expand Up @@ -325,7 +319,6 @@ void TensorRTEngine::FreezeNetwork() {
infer_engine_.reset(infer_builder_->buildEngineWithConfig(
*network(), *infer_builder_config_));
#else
infer_builder_config_->setFlag(nvinfer1::BuilderFlag::kSPARSE_WEIGHTS);
ihost_memory_.reset(infer_builder_->buildSerializedNetwork(
*network(), *infer_builder_config_));
infer_ptr<nvinfer1::IRuntime> runtime(createInferRuntime(&logger_));
Expand Down
4 changes: 4 additions & 0 deletions paddle/fluid/inference/tensorrt/helper.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,6 +96,10 @@ static std::tuple<int, int, int> GetTrtCompileVersion() {
NV_TENSORRT_MAJOR, NV_TENSORRT_MINOR, NV_TENSORRT_PATCH};
}

static float TrtMajorVersion(int full_version) {
return (full_version / 100) / 10.0;
}

template <typename T>
struct Destroyer {
void operator()(T* x) {
Expand Down