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

[inference][trt] update trt hardswish plugin to layer #47745

Merged
merged 2 commits into from
Nov 16, 2022
Merged

[inference][trt] update trt hardswish plugin to layer #47745

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d86dd3b
update hardswish
zhangjun Nov 7, 2022
69c87b1
update
zhangjun Nov 8, 2022
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
43 changes: 26 additions & 17 deletions paddle/fluid/inference/tensorrt/convert/hard_swish_op.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,6 @@ See the License for the specific language governing permissions and
limitations under the License. */

#include "paddle/fluid/inference/tensorrt/convert/op_converter.h"
#include "paddle/fluid/inference/tensorrt/plugin/hard_swish_op_plugin.h"

namespace paddle {
namespace framework {
Expand Down Expand Up @@ -41,7 +40,6 @@ class HardSwishOpConverter : public OpConverter {

framework::OpDesc op_desc(op, nullptr);
// Declare inputs
int input_num = op_desc.Input("X").size();
auto* input = engine_->GetITensor(op_desc.Input("X")[0]);

const float threshold =
Expand Down Expand Up @@ -69,21 +67,32 @@ class HardSwishOpConverter : public OpConverter {
nvinfer1::ElementWiseOperation::kPROD);
layer = eltwise_layer;
} else {
if (engine_->with_dynamic_shape()) {
#if IS_TRT_VERSION_GE(6000)
plugin::HardSwishPluginDynamic* plugin =
new plugin::HardSwishPluginDynamic(threshold, scale, offset);
layer = engine_->AddDynamicPlugin(&input, input_num, plugin);
#else
PADDLE_THROW(platform::errors::Fatal(
"You are running the TRT Dynamic Shape mode, need to confirm that "
"your TRT version is no less than 6.0"));
#endif
} else {
plugin::HardSwishPlugin* plugin =
new plugin::HardSwishPlugin(threshold, scale, offset);
layer = engine_->AddPlugin(&input, input_num, plugin);
}
int32_t rank = input->getDimensions().nbDims;
nvinfer1::Dims constant_shape;
constant_shape.nbDims = rank;
std::fill(constant_shape.d, constant_shape.d + rank, 1);
std::vector<float> weight_threshold_data{threshold};
std::vector<float> weight_scale_data{scale};
std::vector<float> weight_offset_data{offset};
std::vector<float> weight_zero_data{0.f};
auto* threshold_data =
AddConstantLayer(weight_threshold_data.data(), constant_shape);
auto* scale_data =
AddConstantLayer(weight_scale_data.data(), constant_shape);
auto* offset_data =
AddConstantLayer(weight_offset_data.data(), constant_shape);
auto* zero_data =
AddConstantLayer(weight_zero_data.data(), constant_shape);

auto* input_sum_with_offset = Sum(input, offset_data);
auto* pre_max_with_zero = Max(input_sum_with_offset, zero_data);
auto* pre_min_with_threshold = Min(pre_max_with_zero, threshold_data);
auto* pre_prod_with_input = Prod(pre_min_with_threshold, input);
layer = TRT_ENGINE_ADD_LAYER(engine_,
ElementWise,
*pre_prod_with_input,
*scale_data,
nvinfer1::ElementWiseOperation::kDIV);
}
auto output_name = op_desc.Output("Out")[0];
RreplenishLayerAndOutput(layer, "hard_swish", {output_name}, test_mode);
Expand Down