apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-19 06:54:41 +08:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

Merge branch 'develop' of https://github.com/PaddlePaddle/FastDeploy into new_dev

Browse Source
This commit is contained in:
yunyaoXYY
2023-03-21 02:21:35 +00:00
parent f23c61328e f278ed424b
commit 3cc72765dd
53 changed files with 115 additions and 64 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
benchmark/cpp/benchmark_gpu.sh
View File

@@ -21,12 +21,18 @@ fi
./benchmark_ppseg --model Portrait_PP_HumanSegV2_Lite_256x144_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_HumanSegV2_Lite_192x192_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_HumanSegV1_Lite_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_HumanSegV2_Mobile_192x192_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model Deeplabv3_ResNet101_OS8_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH
./benchmark_ppseg --model SegFormer_B0-cityscapes-with-argmax --image cityscapes_demo.png --config_path $CONFIG_PATH
./benchmark_ppmatting --model PP-Matting-512 --image matting_input.jpg --config_path $CONFIG_PATH
./benchmark_ppmatting --model PPHumanMatting --image matting_input.jpg --config_path $CONFIG_PATH
./benchmark_ppmatting --model PPModnet_MobileNetV2 --image matting_input.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model Unet_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_HumanSegV1_Server_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model FCN_HRNet_W18_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH
# PaddleClas
./benchmark_ppcls --model PPLCNet_x1_0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH

29
benchmark/cpp/benchmark_gpu_trt.sh
View File

@@ -21,27 +21,52 @@ fi
./benchmark_ppseg --model Portrait_PP_HumanSegV2_Lite_256x144_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH --trt_shape 1,3,144,256:1,3,144,256:1,3,144,256
./benchmark_ppseg --model PP_HumanSegV2_Lite_192x192_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH --trt_shape 1,3,192,192:1,3,192,192:1,3,192,192
./benchmark_ppseg --model PP_HumanSegV1_Lite_infer --image portrait_heng.jpg --config_path $CONFIG_PATH --trt_shape 1,3,192,192:1,3,192,192:1,3,192,192
./benchmark_ppseg --model PP_HumanSegV2_Mobile_192x192_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH --trt_shape 1,3,192,192:1,3,192,192:1,3,192,192
./benchmark_ppseg --model Deeplabv3_ResNet101_OS8_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH --trt_shape 1,3,512,512:1,3,512,512:1,3,512,512
./benchmark_ppseg --model PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH --trt_shape 1,3,512,512:1,3,512,512:1,3,512,512
./benchmark_ppseg --model SegFormer_B0-cityscapes-with-argmax --image cityscapes_demo.png --config_path $CONFIG_PATH --trt_shape 1,3,512,512:1,3,512,512:1,3,512,512
./benchmark_ppmatting --model PP-Matting-512 --image matting_input.jpg --config_path $CONFIG_PATH --trt_shape 1,3,512,512:1,3,512,512:1,3,512,512
./benchmark_ppmatting --model PPHumanMatting --image matting_input.jpg --config_path $CONFIG_PATH --trt_shape 1,3,2048,2048:1,3,2048,2048:1,3,2048,2048
./benchmark_ppmatting --model PPModnet_MobileNetV2 --image matting_input.jpg --config_path $CONFIG_PATH --trt_shape 1,3,512,512:1,3,512,512:1,3,512,512
./benchmark_ppseg --model Unet_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH --trt_shape 1,3,512,512:1,3,512,512:1,3,512,512
./benchmark_ppseg --model PP_HumanSegV1_Server_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH --trt_shape 1,3,512,512:1,3,512,512:1,3,512,512
./benchmark_ppseg --model FCN_HRNet_W18_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH --trt_shape 1,3,512,512:1,3,512,512:1,3,512,512
# PaddleClas
./benchmark_ppcls --model PPLCNet_x1_0_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
./benchmark_ppcls --model EfficientNetB7_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
./benchmark_ppcls --model GhostNet_x0_5_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
./benchmark_ppcls --model GhostNet_x1_3_ssld_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
./benchmark_ppcls --model MobileNetV1_ssld_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
./benchmark_ppcls --model MobileNetV2_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
./benchmark_ppcls --model MobileNetV3_large_x1_0_ssld_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
./benchmark_ppcls --model ShuffleNetV2_x2_0_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
./benchmark_ppcls --model InceptionV3_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
./benchmark_ppcls --model EfficientNetB0_small_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
./benchmark_ppcls --model PPHGNet_base_ssld_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
./benchmark_ppcls --model ResNet50_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
./benchmark_ppcls --model EfficientNetB0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
./benchmark_ppcls --model MobileNetV2_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
./benchmark_ppcls --model ViT_large_patch16_224_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
./benchmark_ppcls --model DenseNet121_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
./benchmark_ppcls --model person_exists_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH
# 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_ppocr_cls --model ch_ppocr_mobile_v2.0_cls_infer --image rec_img.jpg --config_path $CONFIG_PATH --trt_shape 1,3,48,10:4,3,48,320:8,3,48,1024
./benchmark_ppocr_rec --model ch_PP-OCRv3_rec_infer --image rec_img.jpg --rec_label_file ppocr_keys_v1.txt --config_path $CONFIG_PATH --trt_shape 1,3,48,10:4,3,48,320:8,3,48,2304
./benchmark_ppocr_det --model ch_PP-OCRv2_det_infer --image 12.jpg --config_path $CONFIG_PATH --trt_shape 1,3,64,64:1,3,640,640:1,3,960,960
./benchmark_ppocr_rec --model ch_PP-OCRv2_rec_infer --image rec_img.jpg --rec_label_file ppocr_keys_v1.txt --config_path $CONFIG_PATH --trt_shape 1,3,48,10:4,3,48,320:8,3,48,2304
./benchmark_ppocr_rec --model ch_PP-OCRv2_rec_infer --image rec_img.jpg --rec_label_file ppocr_keys_v1.txt --config_path $CONFIG_PATH --trt_shape 1,3,32,10:4,3,32,320:8,3,32,2304

0
benchmark/cpp/benchmark_picodet.cc Normal file → Executable file
View File

0
benchmark/cpp/benchmark_ppmatting.cc Normal file → Executable file
View File

0
benchmark/cpp/benchmark_ppocr_cls.cc Normal file → Executable file
View File

0
benchmark/cpp/benchmark_ppocr_det.cc Normal file → Executable file
View File

5
benchmark/cpp/benchmark_ppocr_rec.cc Normal file → Executable file
View File

@@ -55,6 +55,11 @@ int main(int argc, char* argv[]) {
}
auto model_ppocr_rec = vision::ocr::Recognizer(
model_file, params_file, FLAGS_rec_label_file, option, model_format);
std::vector<std::string> model_names;
fastdeploy::benchmark::Split(FLAGS_model, model_names, sep);
if (model_names[model_names.size() - 1] == "ch_PP-OCRv2_rec_infer") {
model_ppocr_rec.GetPreprocessor().SetRecImageShape({3, 32, 320});
}
std::string text;
float rec_score;
if (config_info["precision_compare"] == "true") {

0
benchmark/cpp/benchmark_ppseg.cc Normal file → Executable file
View File

0
benchmark/cpp/benchmark_ppyoloe.cc Normal file → Executable file
View File

0
benchmark/cpp/benchmark_ppyolov5.cc Normal file → Executable file
View File

0
benchmark/cpp/benchmark_ppyolov6.cc Normal file → Executable file
View File

0
benchmark/cpp/benchmark_ppyolov7.cc Normal file → Executable file
View File

0
benchmark/cpp/benchmark_ppyolov8.cc Normal file → Executable file
View File

0
benchmark/cpp/benchmark_ppyolox.cc Normal file → Executable file
View File

5
benchmark/cpp/benchmark_x86.sh
View File

@@ -21,12 +21,17 @@ fi
./benchmark_ppseg --model Portrait_PP_HumanSegV2_Lite_256x144_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_HumanSegV2_Lite_192x192_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_HumanSegV1_Lite_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_HumanSegV2_Mobile_192x192_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model Deeplabv3_ResNet101_OS8_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH
./benchmark_ppseg --model SegFormer_B0-cityscapes-with-argmax --image cityscapes_demo.png --config_path $CONFIG_PATH
./benchmark_ppmatting --model PP-Matting-512 --image matting_input.jpg --config_path $CONFIG_PATH
./benchmark_ppmatting --model PPHumanMatting --image matting_input.jpg --config_path $CONFIG_PATH
./benchmark_ppmatting --model PPModnet_MobileNetV2 --image matting_input.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model Unet_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH
./benchmark_ppseg --model PP_HumanSegV1_Server_with_argmax_infer --image portrait_heng.jpg --config_path $CONFIG_PATH
./benchmark_ppseg --model FCN_HRNet_W18_cityscapes_with_argmax_infer --image cityscapes_demo.png --config_path $CONFIG_PATH
# PaddleClas
./benchmark_ppcls --model PPLCNet_x1_0_infer --image ILSVRC2012_val_00000010.jpeg --config_path $CONFIG_PATH

4
benchmark/cpp/config/config.gpu.ort.fp32.e2e.mem.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 10
repeat: 50
warmup: 20
repeat: 100
backend: ort
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.paddle.fp32.e2e.mem.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 10
repeat: 50
warmup: 20
repeat: 100
backend: paddle
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.paddle_trt.fp16.e2e.mem.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 10
repeat: 50
warmup: 20
repeat: 100
backend: paddle_trt
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.paddle_trt.fp16.e2e.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 20
repeat: 100
warmup: 200
repeat: 1000
backend: paddle_trt
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.paddle_trt.fp16.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 20
repeat: 100
warmup: 200
repeat: 1000
backend: paddle_trt
profile_mode: runtime
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.paddle_trt.fp32.e2e.mem.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 10
repeat: 50
warmup: 20
repeat: 100
backend: paddle_trt
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.paddle_trt.fp32.e2e.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 20
repeat: 100
warmup: 200
repeat: 1000
backend: paddle_trt
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.paddle_trt.fp32.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 20
repeat: 100
warmup: 200
repeat: 1000
backend: paddle_trt
profile_mode: runtime
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.trt.fp16.e2e.mem.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 10
repeat: 50
warmup: 20
repeat: 100
backend: trt
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.trt.fp16.e2e.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 20
repeat: 100
warmup: 200
repeat: 1000
backend: trt
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.trt.fp16.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 20
repeat: 100
warmup: 200
repeat: 1000
backend: trt
profile_mode: runtime
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.trt.fp32.e2e.mem.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 10
repeat: 50
warmup: 20
repeat: 100
backend: trt
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.trt.fp32.e2e.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 20
repeat: 100
warmup: 200
repeat: 1000
backend: trt
profile_mode: end2end
include_h2d_d2h: false

4
benchmark/cpp/config/config.gpu.trt.fp32.txt
View File

@@ -1,8 +1,8 @@
device: gpu
device_id: 3
cpu_thread_nums: 1
warmup: 20
repeat: 100
warmup: 200
repeat: 1000
backend: trt
profile_mode: runtime
include_h2d_d2h: false

5
benchmark/cpp/get_models.sh
View File

@@ -152,6 +152,11 @@ download_fd_model_zxvf SegFormer_B0-cityscapes-with-argmax.tgz
download_fd_model_xvf PP-Matting-512.tgz
download_fd_model_xvf PPHumanMatting.tgz
download_fd_model_xvf PPModnet_MobileNetV2.tgz
download_fd_model_xvf Unet_cityscapes_with_argmax_infer.tgz
download_fd_model_xvf PP_HumanSegV1_Server_with_argmax_infer.tgz
download_fd_model_xvf FCN_HRNet_W18_cityscapes_with_argmax_infer.tgz
download_fd_model_xvf PP_HumanSegV2_Mobile_192x192_with_argmax_infer.tgz
# PaddleOCR
download_common_model_xvf https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_det_infer.tar ch_PP-OCRv3_det_infer.tar

2
benchmark/cpp/option.h
View File

@@ -43,6 +43,8 @@ static bool CreateRuntimeOption(fastdeploy::RuntimeOption* option,
option->UsePaddleInferBackend();
} else if (config_info["backend"] == "trt" ||
config_info["backend"] == "paddle_trt") {
option->trt_option.serialize_file = FLAGS_model +
sep + "trt_serialized.trt";
option->UseTrtBackend();
if (config_info["backend"] == "paddle_trt") {
option->UsePaddleInferBackend();

0
benchmark/python/README.md Normal file → Executable file
View File

0
benchmark/python/benchmark_ernie_seq_cls.py Normal file → Executable file
View File

0
benchmark/python/convert_info.py Normal file → Executable file
View File

0
benchmark/python/requirements.txt Normal file → Executable file
View File

0
benchmark/python/run_benchmark_ernie_seq_cls.sh Normal file → Executable file
View File

0
benchmark/python/run_benchmark_ppcls.sh Normal file → Executable file
View File

0
benchmark/python/run_benchmark_ppdet.sh Normal file → Executable file
View File

0
benchmark/python/run_benchmark_ppocr.sh Normal file → Executable file
View File

0
benchmark/python/run_benchmark_ppseg.sh Normal file → Executable file
View File

0
benchmark/python/run_benchmark_uie.sh Normal file → Executable file
View File

16
examples/vision/segmentation/paddleseg/README.md
View File

@@ -30,7 +30,7 @@
|ARM CPU|✅|[链接](semantic_segmentation/cpu-gpu)|✅|✅|
|Intel GPU(集成显卡)|✅|[链接](semantic_segmentation/cpu-gpu)|✅|✅|
|Intel GPU(独立显卡)|✅|[链接](semantic_segmentation/cpu-gpu)|✅|✅|
|昆仑|✅|[链接](semantic_segmentation/kunlun)|✅|✅|
|昆仑|✅|[链接](semantic_segmentation/kunlunxin)|✅|✅|
|昇腾|✅|[链接](semantic_segmentation/ascend)|✅|✅|
|瑞芯微|✅|[链接](semantic_segmentation/rockchip)|✅|✅|
|晶晨|✅|[链接](semantic_segmentation/amlogic)|--|✅|✅|
@@ -58,9 +58,9 @@
- [Python部署示例](semantic_segmentation/cpu-gpu/python/)
- [C++部署示例](semantic_segmentation/cpu-gpu/cpp/)
- 昆仑 XPU
- [部署模型准备](semantic_segmentation/kunlun)
- [Python部署示例](semantic_segmentation/kunlun/python/)
- [C++部署示例](semantic_segmentation/kunlun/cpp/)
- [部署模型准备](semantic_segmentation/kunlunxin)
- [Python部署示例](semantic_segmentation/kunlunxin/python/)
- [C++部署示例](semantic_segmentation/kunlunxin/cpp/)
- 昇腾 Ascend
- [部署模型准备](semantic_segmentation/ascend)
- [Python部署示例](semantic_segmentation/ascend/python/)
@@ -97,7 +97,7 @@
|ARM CPU|✅|[链接](matting/cpu-gpu)|✅|✅|
|Intel GPU(集成显卡)|✅|[链接](matting/cpu-gpu)|✅|✅|
|Intel GPU(独立显卡)|✅|[链接](matting/cpu-gpu)|✅|✅|
|昆仑|✅|[链接](matting/kunlun)|✅|✅|
|昆仑|✅|[链接](matting/kunlunxin)|✅|✅|
|昇腾|✅|[链接](matting/ascend)|✅|✅|
### 3.2 详细使用文档
@@ -122,9 +122,9 @@
- [Python部署示例](matting/cpu-gpu/python/)
- [C++部署示例](cpu-gpu/cpp/)
- 昆仑 XPU
- [部署模型准备](matting/kunlun)
- [Python部署示例](matting/kunlun/README.md)
- [C++部署示例](matting/kunlun/README.md)
- [部署模型准备](matting/kunlunxin)
- [Python部署示例](matting/kunlunxin/README.md)
- [C++部署示例](matting/kunlunxin/README.md)
- 昇腾 Ascend
- [部署模型准备](matting/ascend)
- [Python部署示例](matting/ascend/README.md)

8
examples/vision/segmentation/paddleseg/matting/README.md
View File

@@ -13,7 +13,7 @@
|ARM CPU|✅|[链接](cpu-gpu)|✅|✅|
|Intel GPU(集成显卡)|✅|[链接](cpu-gpu)|✅|✅|
|Intel GPU(独立显卡)|✅|[链接](cpu-gpu)|✅|✅|
|昆仑|✅|[链接](kunlun)|✅|✅|
|昆仑|✅|[链接](kunlunxin)|✅|✅|
|昇腾|✅|[链接](ascend)|✅|✅|
## 3. 详细使用文档
@@ -38,9 +38,9 @@
- [Python部署示例](cpu-gpu/python/)
- [C++部署示例](cpu-gpu/cpp/)
- 昆仑 XPU
- [部署模型准备](kunlun)
- [Python部署示例](kunlun/README.md)
- [C++部署示例](kunlun/README.md)
- [部署模型准备](kunlunxin)
- [Python部署示例](kunlunxin/README.md)
- [C++部署示例](kunlunxin/README.md)
- 昇腾 Ascend
- [部署模型准备](ascend)
- [Python部署示例](ascend/README.md)

0
examples/vision/segmentation/paddleseg/matting/kunlun/README.md → examples/vision/segmentation/paddleseg/matting/kunlunxin/README.md
View File

8
examples/vision/segmentation/paddleseg/semantic_segmentation/README.md
View File

@@ -13,7 +13,7 @@
|ARM CPU|✅|[链接](cpu-gpu)|✅|✅|
|Intel GPU(集成显卡)|✅|[链接](cpu-gpu)|✅|✅|
|Intel GPU(独立显卡)|✅|[链接](cpu-gpu)|✅|✅|
|昆仑|✅|[链接](kunlun)|✅|✅|
|昆仑|✅|[链接](kunlunxin)|✅|✅|
|昇腾|✅|[链接](ascend)|✅|✅|
|瑞芯微|✅|[链接](rockchip)|✅|✅|
|晶晨|✅|[链接](amlogic)|--|✅|
@@ -41,9 +41,9 @@
- [Python部署示例](cpu-gpu/python/)
- [C++部署示例](cpu-gpu/cpp/)
- 昆仑 XPU
- [部署模型准备](kunlun)
- [Python部署示例](kunlun/python/)
- [C++部署示例](kunlun/cpp/)
- [部署模型准备](kunlunxin)
- [Python部署示例](kunlunxin/python/)
- [C++部署示例](kunlunxin/cpp/)
- 昇腾 Ascend
- [部署模型准备](ascend)
- [Python部署示例](ascend/python/)

0
examples/vision/segmentation/paddleseg/semantic_segmentation/kunlun/README.md → examples/vision/segmentation/paddleseg/semantic_segmentation/kunlunxin/README.md
View File

0
examples/vision/segmentation/paddleseg/semantic_segmentation/kunlun/cpp/CMakeLists.txt → examples/vision/segmentation/paddleseg/semantic_segmentation/kunlunxin/cpp/CMakeLists.txt
View File

4
examples/vision/segmentation/paddleseg/semantic_segmentation/kunlun/cpp/README.md → examples/vision/segmentation/paddleseg/semantic_segmentation/kunlunxin/cpp/README.md
View File

@@ -14,12 +14,12 @@
```bash
# 下载部署示例代码
git clone https://github.com/PaddlePaddle/FastDeploy.git
cd FastDeploy/examples/vision/segmentation/semantic_segmentation/kunlun/cpp
cd FastDeploy/examples/vision/segmentation/semantic_segmentation/kunlunxin/cpp
# 如果您希望从PaddleSeg下载示例代码，请运行
# git clone https://github.com/PaddlePaddle/PaddleSeg.git
# # 注意：如果当前分支找不到下面的fastdeploy测试代码，请切换到develop分支
# # git checkout develop
# cd PaddleSeg/deploy/fastdeploy/semantic_segmentation/kunlun/cpp
# cd PaddleSeg/deploy/fastdeploy/semantic_segmentation/kunlunxin/cpp
mkdir build
cd build

0
examples/vision/segmentation/paddleseg/semantic_segmentation/kunlun/cpp/infer.cc → examples/vision/segmentation/paddleseg/semantic_segmentation/kunlunxin/cpp/infer.cc
View File

4
examples/vision/segmentation/paddleseg/semantic_segmentation/kunlun/python/README.md → examples/vision/segmentation/paddleseg/semantic_segmentation/kunlunxin/python/README.md
View File

@@ -14,12 +14,12 @@
```bash
# 下载部署示例代码
git clone https://github.com/PaddlePaddle/FastDeploy.git
cd FastDeploy/examples/vision/segmentation/semantic_segmentation/kunlun/python
cd FastDeploy/examples/vision/segmentation/semantic_segmentation/kunlunxin/python
# 如果您希望从PaddleSeg下载示例代码，请运行
# git clone https://github.com/PaddlePaddle/PaddleSeg.git
# # 注意：如果当前分支找不到下面的fastdeploy测试代码，请切换到develop分支
# # git checkout develop
# cd PaddleSeg/deploy/fastdeploy/semantic_segmentation/kunlun/python
# cd PaddleSeg/deploy/fastdeploy/semantic_segmentation/kunlunxin/python
# 下载PP-LiteSeg模型文件和测试图片
wget https://bj.bcebos.com/paddlehub/fastdeploy/PP_LiteSeg_B_STDC2_cityscapes_without_argmax_infer.tgz

0
examples/vision/segmentation/paddleseg/semantic_segmentation/kunlun/python/infer.py → examples/vision/segmentation/paddleseg/semantic_segmentation/kunlunxin/python/infer.py
View File

25
fastdeploy/runtime/backends/paddle/paddle_backend.cc
View File

@@ -113,7 +113,8 @@ bool PaddleBackend::Init(const RuntimeOption& runtime_option) {
option.paddle_infer_option.external_stream_ = runtime_option.external_stream_;
option.paddle_infer_option.trt_option = runtime_option.trt_option;
option.paddle_infer_option.trt_option.gpu_id = runtime_option.device_id;
return InitFromPaddle(option.model_file, option.params_file, option.model_from_memory_, option.paddle_infer_option);
return InitFromPaddle(option.model_file, option.params_file,
option.model_from_memory_, option.paddle_infer_option);
}
bool PaddleBackend::InitFromPaddle(const std::string& model,
@@ -126,8 +127,8 @@ bool PaddleBackend::InitFromPaddle(const std::string& model,
return false;
}
if (model_from_memory) {
config_.SetModelBuffer(model.c_str(), model.size(),
params.c_str(), params.size());
config_.SetModelBuffer(model.c_str(), model.size(), params.c_str(),
params.size());
} else {
config_.SetModel(model, params);
}
@@ -140,7 +141,8 @@ bool PaddleBackend::InitFromPaddle(const std::string& model,
// PaddleReader instead now
std::string model_content = model;
if (!model_from_memory) {
FDASSERT(ReadBinaryFromFile(model, &model_content), "Failed to read file %s.", model.c_str());
FDASSERT(ReadBinaryFromFile(model, &model_content),
"Failed to read file %s.", model.c_str());
}
auto reader =
paddle2onnx::PaddleReader(model_content.c_str(), model_content.size());
@@ -210,8 +212,7 @@ bool PaddleBackend::InitFromPaddle(const std::string& model,
paddle_infer::Config analysis_config;
if (model_from_memory) {
analysis_config.SetModelBuffer(model.c_str(), model.size(),
params.c_str(),
params.size());
params.c_str(), params.size());
} else {
analysis_config.SetModel(model, params);
}
@@ -283,7 +284,6 @@ bool PaddleBackend::Infer(std::vector<FDTensor>& inputs,
auto handle = predictor_->GetInputHandle(inputs[i].name);
ShareTensorFromFDTensor(handle.get(), inputs[i]);
}
std::unordered_set<std::string> prebinded_output_name;
// prebinded output only support for GPU
if (!copy_to_fd) {
for (size_t i = 0; i < (*outputs).size(); ++i) {
@@ -297,7 +297,6 @@ bool PaddleBackend::Infer(std::vector<FDTensor>& inputs,
// Record the prebinded output_name.
// Those outputs do not need PaddleTensorToFDTensor
// after predictor_.Run()
prebinded_output_name.insert(output_name);
auto handle = predictor_->GetOutputHandle(output_name);
ShareOutTensorFromFDTensor(handle.get(), (*outputs)[i]);
}
@@ -309,11 +308,6 @@ bool PaddleBackend::Infer(std::vector<FDTensor>& inputs,
outputs->resize(outputs_desc_.size());
for (size_t i = 0; i < outputs_desc_.size(); ++i) {
// skip prebinded output
if (copy_to_fd == false &&
prebinded_output_name.count(outputs_desc_[i].name)) {
continue;
}
auto handle = predictor_->GetOutputHandle(outputs_desc_[i].name);
if (copy_to_fd) {
(*outputs)[i].is_pinned_memory = option_.enable_pinned_memory;
@@ -334,7 +328,10 @@ std::unique_ptr<BaseBackend> PaddleBackend::Clone(RuntimeOption& runtime_option,
auto clone_option = option_;
clone_option.device_id = device_id;
clone_option.external_stream_ = stream;
FDASSERT(casted_backend->InitFromPaddle(runtime_option.model_file, runtime_option.params_file, runtime_option.model_from_memory_, clone_option), "Clone model from Paddle failed while initialize PaddleBackend.");
FDASSERT(casted_backend->InitFromPaddle(
runtime_option.model_file, runtime_option.params_file,
runtime_option.model_from_memory_, clone_option),
"Clone model from Paddle failed while initialize PaddleBackend.");
FDWARNING << "The target device id:" << device_id
<< " is different from current device id:" << option_.device_id
<< ", cannot share memory with current engine." << std::endl;

6
serving/src/fastdeploy_runtime.cc
View File

@@ -345,6 +345,12 @@ ModelState::ModelState(TRITONBACKEND_Model* triton_model)
ParseBoolValue(value_string, &enable_fixed_size_opt));
runtime_options_->paddle_infer_option.enable_fixed_size_opt =
enable_fixed_size_opt;
} else if (param_key == "collect_trt_shape") {
bool collect_trt_shape = false;
THROW_IF_BACKEND_MODEL_ERROR(
ParseBoolValue(value_string, &collect_trt_shape));
runtime_options_->paddle_infer_option.collect_trt_shape =
collect_trt_shape;
}
}
}