Skip to content
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

Jump to bottom

[Paddle-TRT] fix skiplayernorm, add trt_version check #52342

Merged
merged 4 commits into from
Mar 31, 2023
Merged

[Paddle-TRT] fix skiplayernorm, add trt_version check #52342

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
2ded0bf
fix skiplayernorm, add trt_version check
Wangzheee Mar 30, 2023
56c48a3
fix
Wangzheee Mar 30, 2023
3137d7e
fix
Wangzheee Mar 30, 2023
3628f91
fix
Wangzheee Mar 30, 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
17 changes: 17 additions & 0 deletions paddle/fluid/framework/ir/trt_skip_layernorm_fuse_pass.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,9 @@ limitations under the License. */

#include "paddle/fluid/framework/ir/graph_pattern_detector.h"
#include "paddle/fluid/framework/op_version_registry.h"
#ifdef PADDLE_WITH_TENSORRT
#include "paddle/fluid/inference/tensorrt/helper.h"
#endif

namespace paddle {
namespace framework {
Expand Down Expand Up @@ -105,6 +108,20 @@ void TrtSkipLayerNormFusePass::ApplyImpl(ir::Graph *graph) const {
PADDLE_ENFORCE_NOT_NULL(
graph, platform::errors::PreconditionNotMet("graph should not be null."));
FusePassBase::Init("skip_layernorm_fuse", graph);

#ifdef PADDLE_WITH_TENSORRT
auto trt_version = paddle::inference::tensorrt::GetTrtRuntimeVersion();
if (std::get<0>(trt_version) * 1000 + std::get<1>(trt_version) * 100 +
std::get<2>(trt_version) * 10 <
7200) {
VLOG(3) << "skip_layernorm oss plugin only available for trt version >= "
"7.2 Stop this pass";
return;
}
#else
// if no tensorrt, early stop
return;
#endif
int found_subgraph_count = 0;

GraphPatternDetector gpd;
Expand Down