mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-12-24 13:28:13 +08:00
* add ppcls benchmark * add ppcls benchmark * add ppcls benchmark * add ppcls benchmark * fixed txt path * resolve conflict * resolve conflict * deal with comments * Add enable_trt_fp16 option * Add OV backend for seg and det * fixed valid backends in ppdet * fixed valid backends in yolo * add copyright and rm Chinese Notes * add ppdet&ppseg&yolo benchmark * add cpu/gpu mem info * Add benchmark readme * fixed bug Co-authored-by: Jason <jiangjiajun@baidu.com>
92 lines
3.7 KiB
Markdown
92 lines
3.7 KiB
Markdown
# FastDeploy Benchmarks
|
||
|
||
在跑benchmark前,需确认以下两个步骤
|
||
|
||
* 1. 软硬件环境满足要求,参考[FastDeploy环境要求](../docs/environment.md)
|
||
* 2. FastDeploy Python whl包安装,参考[FastDeploy Python安装](../docs/quick_start)
|
||
|
||
FastDeploy 目前支持多种推理后端,下面以 PaddleClas MobileNetV1 为例,跑出多后端在 CPU/GPU 对应 benchmark 数据
|
||
|
||
```bash
|
||
# 下载 MobileNetV1 模型
|
||
wget https://bj.bcebos.com/paddlehub/fastdeploy/MobileNetV1_x0_25_infer.tgz
|
||
tar -xvf MobileNetV1_x0_25_infer
|
||
|
||
# 下载图片
|
||
wget https://gitee.com/paddlepaddle/PaddleClas/raw/release/2.4/deploy/images/ImageNet/ILSVRC2012_val_00000010.jpeg
|
||
|
||
# CPU
|
||
# Paddle Inference
|
||
python benchmark_ppcls.py --model MobileNetV1_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --cpu_num_thread 8 --iter_num 2000 --backend paddle
|
||
|
||
# ONNX Runtime
|
||
python benchmark_ppcls.py --model MobileNetV1_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --cpu_num_thread 8 --iter_num 2000 --backend ort
|
||
|
||
# OpenVINO
|
||
python benchmark_ppcls.py --model MobileNetV1_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --cpu_num_thread 8 --iter_num 2000 --backend ov
|
||
|
||
# GPU
|
||
# Paddle Inference
|
||
python benchmark_ppcls.py --model MobileNetV1_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --device gpu --iter_num 2000 --backend paddle
|
||
|
||
# ONNX Runtime
|
||
python benchmark_ppcls.py --model MobileNetV1_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --device gpu --iter_num 2000 --backend ort
|
||
|
||
# TensorRT
|
||
python benchmark_ppcls.py --model MobileNetV1_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --device gpu --iter_num 2000 --backend trt
|
||
|
||
# TensorRT fp16
|
||
python benchmark_ppcls.py --model MobileNetV1_x0_25_infer --image ILSVRC2012_val_00000010.jpeg --device gpu --iter_num 2000 --backend trt --enable_trt_fp16 True
|
||
|
||
```
|
||
|
||
**具体参数说明**
|
||
|
||
| 参数 | 作用 |
|
||
| -------------------- | ------------------------------------------ |
|
||
| --model | 模型路径 |
|
||
| --image | 图片路径 |
|
||
| --device | 选择 CPU 还是 GPU,默认 CPU |
|
||
| --cpu_num_thread | CPU 线程数 |
|
||
| --device_id | GPU 卡号 |
|
||
| --iter_num | 跑 benchmark 的迭代次数 |
|
||
| --backend | 指定后端类型,有ort, ov, trt, paddle四个选项 |
|
||
| --enable_trt_fp16 | 当后端为trt时,是否开启fp16 |
|
||
|
||
**最终txt结果**
|
||
|
||
将当前目录的所有txt汇总并结构化,执行下列命令
|
||
|
||
```bash
|
||
# 汇总
|
||
cat *.txt >> ./result_ppcls.txt
|
||
|
||
# 结构化信息
|
||
python convert_info.py --txt_path result_ppcls.txt --domain ppcls
|
||
```
|
||
|
||
得到 CPU 结果```struct_cpu_ppcls.txt```以及 GPU 结果```struct_gpu_ppcls.txt```如下所示
|
||
|
||
```bash
|
||
# struct_cpu_ppcls.txt
|
||
model_name thread_nums ort_run ort_end2end cpu_rss_mb ov_run ov_end2end cpu_rss_mb paddle_run paddle_end2end cpu_rss_mb
|
||
MobileNetV1_x0_25 8 1.18 3.27 270.43 0.87 1.98 272.26 3.13 5.29 899.57
|
||
|
||
# struct_gpu_ppcls.txt
|
||
model_name ort_run ort_end2end gpu_rss_mb paddle_run paddle_end2end gpu_rss_mb trt_run trt_end2end gpu_rss_mb trt_fp16_run trt_fp16_end2end gpu_rss_mb
|
||
MobileNetV1_x0_25 1.25 3.24 677.06 2.00 3.77 945.06 0.67 2.66 851.06 0.53 2.46 839.06
|
||
```
|
||
|
||
**结果说明**
|
||
|
||
* ```_run```后缀代表一次infer耗时,包括H2D以及D2H;```_end2end```后缀代表包含前后处理耗时
|
||
* ```cpu_rss_mb```代表内存占用;```gpu_rss_mb```代表显存占用
|
||
|
||
若有多个PaddleClas模型,在当前目录新建ppcls_model目录,将所有模型放入该目录即可,运行下列命令
|
||
|
||
```bash
|
||
sh run_benchmark_ppcls.sh
|
||
```
|
||
|
||
一键得到所有模型在 CPU 以及 GPU 的 benchmark 数据
|