[Benchmark] Add dynamic shape for pp_trt (#1703)

* add GPL lisence

* add GPL-3.0 lisence

* add GPL-3.0 lisence

* add GPL-3.0 lisence

* support yolov8

* add pybind for yolov8

* add yolov8 readme

* add cpp benchmark

* add cpu and gpu mem

* public part split

* add runtime mode

* fixed bugs

* add cpu_thread_nums

* deal with comments

* deal with comments

* deal with comments

* rm useless code

* add FASTDEPLOY_DECL

* add FASTDEPLOY_DECL

* fixed for windows

* mv rss to pss

* mv rss to pss

* Update utils.cc

* use thread to collect mem

* Add ResourceUsageMonitor

* rm useless code

* fixed bug

* fixed typo

* update ResourceUsageMonitor

* fixed bug

* fixed bug

* add note for ResourceUsageMonitor

* deal with comments

* add macros

* deal with comments

* deal with comments

* deal with comments

* re-lint

* rm pmap and use mem api

* rm pmap and use mem api

* add mem api

* Add PrintBenchmarkInfo func

* Add PrintBenchmarkInfo func

* Add PrintBenchmarkInfo func

* deal with comments

* fixed enable_paddle_to_trt

* add log for paddle_trt

* support ppcls benchmark

* use new trt option api

* update benchmark info

* simplify benchmark.cc

* simplify benchmark.cc

* deal with comments

* Add ppseg && ppocr benchmark

* add OCR rec img

* add ocr benchmark

* fixed trt shape

* add trt shape

* resolve conflict

* add ENABLE_BENCHMARK define

* Add ClassifyDiff

* Add Resize for ClassifyResult

* deal with comments

* add convert info script

* resolve conflict

* Add SaveBenchmarkResult func

* fixed bug

* fixed bug

* fixed bug

* add config.txt for option

* fixed bug

* fixed bug

* fixed bug

* add benchmark.sh

* mv thread_nums from 8 to 1

* deal with comments

* deal with comments

* fixed readme

* deal with comments

* add all platform shell

* Update config.arm.txt

* Update config.gpu.txt

* Update config.x86.txt

* fixed printinfo bug

* rm proxy

* add more model support

* all backend config.txt

* deal with comments

* Add MattingDiff compare

* fixed predict bug

* adjust warmup/repeat times

* add e2e/mem configs

* fixed typo

* open collect_mem

* fixed typo

* add trt cache option

* fixed bug

* fixed repeat times

* test for benchmark

* test for det benchmark

* for benchmark

* fixed for x86

* add h2d and d2h config

* renmae txt file

* add dynamic shape for pp_trt

* fixed typo

* Update option.h

* add collect shape

---------

Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com>

benchmark/cpp/benchmark_gpu_trt.sh

@@ -8,35 +8,35 @@ if [ ! "$1" = "$CONFIG_PATH" ]; then
 fi
 
 # PaddleClas
-./benchmark_ppcls --model PPLCNet_x1_0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model PPLCNet_x1_0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "x"
-./benchmark_ppcls --model PPLCNetV2_base_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model PPLCNetV2_base_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "x"
-./benchmark_ppcls --model EfficientNetB7_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model EfficientNetB7_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,600,600:1,3,600,600:1,3,600,600" --input_name "x"
-./benchmark_ppcls --model EfficientNetB0_small_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model EfficientNetB0_small_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model GhostNet_x0_5_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model GhostNet_x0_5_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model GhostNet_x1_3_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model GhostNet_x1_3_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "x"
-./benchmark_ppcls --model GhostNet_x1_3_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model GhostNet_x1_3_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model MobileNetV1_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model MobileNetV1_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model MobileNetV1_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model MobileNetV1_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model MobileNetV2_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model MobileNetV2_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model MobileNetV2_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model MobileNetV2_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model MobileNetV3_small_x0_35_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model MobileNetV3_small_x0_35_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "x"
-./benchmark_ppcls --model MobileNetV3_large_x1_0_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model MobileNetV3_large_x1_0_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model ShuffleNetV2_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model ShuffleNetV2_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model ShuffleNetV2_x2_0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model ShuffleNetV2_x2_0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model SqueezeNet1_1_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model SqueezeNet1_1_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model InceptionV3_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model InceptionV3_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,299,299:1,3,299,299:1,3,299,299" --input_name "x"
-./benchmark_ppcls --model ResNet50_vd_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model ResNet50_vd_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model PPHGNet_tiny_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model PPHGNet_tiny_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "x"
-./benchmark_ppcls --model PPHGNet_base_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model PPHGNet_base_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "x"
-./benchmark_ppcls --model ResNet50_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model ResNet50_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model EfficientNetB0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model EfficientNetB0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "x"
-./benchmark_ppcls --model MobileNetV2_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model MobileNetV2_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model MobileNetV3_small_x1_0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model MobileNetV3_small_x1_0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model ViT_large_patch16_224_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model ViT_large_patch16_224_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model ResNeXt50_32x4d_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model ResNeXt50_32x4d_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model DenseNet121_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model DenseNet121_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "inputs"
-./benchmark_ppcls --model PPHGNet_small_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model PPHGNet_small_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "x"
-./benchmark_ppcls --model person_exists_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
+./benchmark_ppcls --model person_exists_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH --trt_shape "1,3,224,224:1,3,224,224:1,3,224,224" --input_name "x"
 
 # PaddleOCR
 ./benchmark_ppocr_det --model ch_PP-OCRv3_det_infer --image 12.jpg --config_path $CONFIG_PATH --trt_shape 1,3,64,64:1,3,640,640:1,3,960,960

benchmark/cpp/benchmark_picodet.cc

@@ -42,6 +42,16 @@ int main(int argc, char* argv[]) {
   auto model_file = FLAGS_model + sep + model_name;
   auto params_file = FLAGS_model + sep + params_name;
   auto config_file = FLAGS_model + sep + config_name;
+  if (config_info["backend"] == "paddle_trt") {
+    option.paddle_infer_option.collect_trt_shape = true;
+  }
+  if (config_info["backend"] == "paddle_trt" ||
+      config_info["backend"] == "trt") {
+    option.trt_option.SetShape("image", {1, 3, 640, 640}, {1, 3, 640, 640},
+                               {1, 3, 640, 640});
+    option.trt_option.SetShape("scale_factor", {1, 2}, {1, 2},
+                               {1, 2});
+  }
   auto model_picodet = vision::detection::PicoDet(
       model_file, params_file, config_file, option, model_format);
   if (FLAGS_no_nms) {

benchmark/cpp/benchmark_ppcls.cc

@@ -19,6 +19,14 @@
 namespace vision = fastdeploy::vision;
 namespace benchmark = fastdeploy::benchmark;
 
+DEFINE_string(trt_shape, "1,3,224,224:1,3,224,224:1,3,224,224",
+              "Set min/opt/max shape for trt/paddle_trt backend."
+              "eg:--trt_shape 1,3,224,224:1,3,224,224:1,3,224,224");
+ 
+DEFINE_string(input_name, "x",
+              "Set input name for trt/paddle_trt backend."
+              "eg:--input_names x");
+ 
 int main(int argc, char* argv[]) {
 #if defined(ENABLE_BENCHMARK) && defined(ENABLE_VISION)
   // Initialization
@@ -44,6 +52,17 @@ int main(int argc, char* argv[]) {
   auto model_file = FLAGS_model + sep + model_name;
   auto params_file = FLAGS_model + sep + params_name;
   auto config_file = FLAGS_model + sep + config_name;
+  if (config_info["backend"] == "paddle_trt") {
+    option.paddle_infer_option.collect_trt_shape = true;
+  }
+  if (config_info["backend"] == "paddle_trt" ||
+      config_info["backend"] == "trt") {
+    std::vector<std::vector<int32_t>> trt_shapes =
+        benchmark::ResultManager::GetInputShapes(FLAGS_trt_shape);
+    option.trt_option.SetShape(FLAGS_input_name, trt_shapes[0], 
+                               trt_shapes[1], trt_shapes[2]);
+  }
+ 
   auto model_ppcls = vision::classification::PaddleClasModel(
       model_file, params_file, config_file, option, model_format);
   vision::ClassifyResult res;

benchmark/cpp/benchmark_ppyoloe.cc

@@ -41,6 +41,16 @@ int main(int argc, char* argv[]) {
   auto model_file = FLAGS_model + sep + model_name;
   auto params_file = FLAGS_model + sep + params_name;
   auto config_file = FLAGS_model + sep + config_name;
+  if (config_info["backend"] == "paddle_trt") {
+    option.paddle_infer_option.collect_trt_shape = true;
+  }
+  if (config_info["backend"] == "paddle_trt" ||
+      config_info["backend"] == "trt") {
+    option.trt_option.SetShape("image", {1, 3, 640, 640}, {1, 3, 640, 640},
+                               {1, 3, 640, 640});
+    option.trt_option.SetShape("scale_factor", {1, 2}, {1, 2},
+                               {1, 2});
+  }
   auto model_ppyoloe = vision::detection::PPYOLOE(
       model_file, params_file, config_file, option, model_format);
   if (FLAGS_no_nms) {

benchmark/cpp/benchmark_ppyolov5.cc

@@ -41,6 +41,16 @@ int main(int argc, char* argv[]) {
   auto model_file = FLAGS_model + sep + model_name;
   auto params_file = FLAGS_model + sep + params_name;
   auto config_file = FLAGS_model + sep + config_name;
+  if (config_info["backend"] == "paddle_trt") {
+    option.paddle_infer_option.collect_trt_shape = true;
+  }
+  if (config_info["backend"] == "paddle_trt" ||
+      config_info["backend"] == "trt") {
+    option.trt_option.SetShape("image", {1, 3, 640, 640}, {1, 3, 640, 640},
+                               {1, 3, 640, 640});
+    option.trt_option.SetShape("scale_factor", {1, 2}, {1, 2},
+                               {1, 2});
+  }
   auto model_ppyolov5 = vision::detection::PaddleYOLOv5(
       model_file, params_file, config_file, option, model_format);
   vision::DetectionResult res;

benchmark/cpp/benchmark_ppyolov6.cc

@@ -41,6 +41,16 @@ int main(int argc, char* argv[]) {
   auto model_file = FLAGS_model + sep + model_name;
   auto params_file = FLAGS_model + sep + params_name;
   auto config_file = FLAGS_model + sep + config_name;
+  if (config_info["backend"] == "paddle_trt") {
+    option.paddle_infer_option.collect_trt_shape = true;
+  }
+  if (config_info["backend"] == "paddle_trt" ||
+      config_info["backend"] == "trt") {
+    option.trt_option.SetShape("image", {1, 3, 640, 640}, {1, 3, 640, 640},
+                               {1, 3, 640, 640});
+    option.trt_option.SetShape("scale_factor", {1, 2}, {1, 2},
+                               {1, 2});
+  }
   auto model_ppyolov6 = vision::detection::PaddleYOLOv6(
       model_file, params_file, config_file, option, model_format);
   vision::DetectionResult res;

benchmark/cpp/benchmark_ppyolov7.cc

@@ -41,6 +41,16 @@ int main(int argc, char* argv[]) {
   auto model_file = FLAGS_model + sep + model_name;
   auto params_file = FLAGS_model + sep + params_name;
   auto config_file = FLAGS_model + sep + config_name;
+  if (config_info["backend"] == "paddle_trt") {
+    option.paddle_infer_option.collect_trt_shape = true;
+  }
+  if (config_info["backend"] == "paddle_trt" ||
+      config_info["backend"] == "trt") {
+    option.trt_option.SetShape("image", {1, 3, 640, 640}, {1, 3, 640, 640},
+                               {1, 3, 640, 640});
+    option.trt_option.SetShape("scale_factor", {1, 2}, {1, 2},
+                               {1, 2});
+  }
   auto model_ppyolov7 = vision::detection::PaddleYOLOv7(
       model_file, params_file, config_file, option, model_format);
   vision::DetectionResult res;

benchmark/cpp/benchmark_ppyolov8.cc

@@ -41,6 +41,16 @@ int main(int argc, char* argv[]) {
   auto model_file = FLAGS_model + sep + model_name;
   auto params_file = FLAGS_model + sep + params_name;
   auto config_file = FLAGS_model + sep + config_name;
+  if (config_info["backend"] == "paddle_trt") {
+    option.paddle_infer_option.collect_trt_shape = true;
+  }
+  if (config_info["backend"] == "paddle_trt" ||
+      config_info["backend"] == "trt") {
+    option.trt_option.SetShape("image", {1, 3, 640, 640}, {1, 3, 640, 640},
+                               {1, 3, 640, 640});
+    option.trt_option.SetShape("scale_factor", {1, 2}, {1, 2},
+                               {1, 2});
+  }
   auto model_ppyolov8 = vision::detection::PaddleYOLOv8(
       model_file, params_file, config_file, option, model_format);
   vision::DetectionResult res;

benchmark/cpp/benchmark_ppyolox.cc

@@ -41,6 +41,16 @@ int main(int argc, char* argv[]) {
   auto model_file = FLAGS_model + sep + model_name;
   auto params_file = FLAGS_model + sep + params_name;
   auto config_file = FLAGS_model + sep + config_name;
+  if (config_info["backend"] == "paddle_trt") {
+    option.paddle_infer_option.collect_trt_shape = true;
+  }
+  if (config_info["backend"] == "paddle_trt" ||
+      config_info["backend"] == "trt") {
+    option.trt_option.SetShape("image", {1, 3, 640, 640}, {1, 3, 640, 640},
+                               {1, 3, 640, 640});
+    option.trt_option.SetShape("scale_factor", {1, 2}, {1, 2},
+                               {1, 2});
+  }
   auto model_ppyolox = vision::detection::PaddleYOLOX(
       model_file, params_file, config_file, option, model_format);
   vision::DetectionResult res;