update auto_log install command (#5482)

* update auto_log install command

* add benchmark output statement

* update benchmark=True and add benchmark declaration

* add train_type prefix of inference log

* update test_train_inference_python.md

* rm prefix of inference log

* add O2 declaration(not support now)

* reset norm train script and tipc doc

tutorials/mobilenetv3_prod/Step6/test_tipc/configs/mobilenet_v3_small/train_amp_infer_python.txt

@@ -29,4 +29,4 @@ inference:deploy/inference_python/infer.py
 --batch-size:1
 --model-dir:./output/mobilenet_v3_small_infer/
 --img-path:./images/demo.jpg
---benchmark:False
\ No newline at end of file
+--benchmark:True
\ No newline at end of file

tutorials/mobilenetv3_prod/Step6/test_tipc/docs/test_train_amp_inference_python.md

@@ -42,16 +42,11 @@ tar -xf test_images/lite_data.tar
 
 - 安装依赖
     ```
-    pip3 install  -r requirements.txt
+    pip install  -r requirements.txt
     ```
 - 安装AutoLog（规范化日志输出工具）
     ```
-    git clone https://github.com/LDOUBLEV/AutoLog
-    cd AutoLog
-    pip install -r requirements.txt
-    python setup.py bdist_wheel
-    pip3 install ./dist/auto_log-1.0.0-py3-none-any.whl
-    cd ../
+    pip install  https://paddleocr.bj.bcebos.com/libs/auto_log-1.2.0-py3-none-any.whl
     ```
 
 ### 2.3 功能测试
@@ -76,9 +71,45 @@ bash test_tipc/test_train_inference_python.sh test_tipc/configs/mobilenet_v3_sma
 输出结果如下，表示命令运行成功。
 
 ```bash
-Run successfully with command - python3.7 train.py --amp_level=O1 --output-dir=./log/mobilenet_v3_small/lite_train_lite_infer/norm_train_gpus_0 --epochs=5   --batch-size=4!
+Run successfully with command - python3.7 train.py --amp_level=O1 --output-dir=./log/mobilenet_v3_small/lite_train_lite_infer/amp_train_gpus_0 --epochs=5   --batch-size=4!
 ......
-Run successfully with command - python3.7 deploy/inference_python/infer.py --use-gpu=False --model-dir=./log/mobilenet_v3_small/lite_train_lite_infer/norm_train_gpus_0,1 --batch-size=1   --benchmark=False > ./log/mobilenet_v3_small/lite_train_lite_infer/python_infer_cpu_batchsize_1.log 2>&1 !
+Run successfully with command - python3.7 deploy/inference_python/infer.py --use-gpu=False --model-dir=./log/mobilenet_v3_small/lite_train_lite_infer/amp_train_gpus_0,1 --batch-size=1   --benchmark=True > ./log/mobilenet_v3_small/lite_train_lite_infer/python_infer_cpu_batchsize_1.log 2>&1 !
 ```
 
+在开启benchmark选项时，可以得到测试的详细数据，包含运行环境信息（系统版本、CUDA版本、CUDNN版本、驱动版本），Paddle版本信息，参数设置信息（运行设备、线程数、是否开启内存优化等），模型信息（模型名称、精度），数据信息（batchsize、是否为动态shape等），性能信息（CPU/GPU的占用、运行耗时、预处理耗时、推理耗时、后处理耗时），内容如下所示：
+
+```
+[2022/03/03 04:21:20] root INFO: ---------------------- Env info ----------------------
+[2022/03/03 04:21:20] root INFO:  OS_version: Ubuntu 16.04
+[2022/03/03 04:21:20] root INFO:  CUDA_version: 10.2.89
+[2022/03/03 04:21:20] root INFO:  CUDNN_version: 7.6.5
+[2022/03/03 04:21:20] root INFO:  drivier_version: 440.64.00
+[2022/03/03 04:21:20] root INFO: ---------------------- Paddle info ----------------------
+[2022/03/03 04:21:20] root INFO:  paddle_version: 2.2.2
+[2022/03/03 04:21:20] root INFO:  paddle_commit: b031c389938bfa15e15bb20494c76f86289d77b0
+[2022/03/03 04:21:20] root INFO:  log_api_version: 1.0
+[2022/03/03 04:21:20] root INFO: ----------------------- Conf info -----------------------
+[2022/03/03 04:21:20] root INFO:  runtime_device: cpu
+[2022/03/03 04:21:20] root INFO:  ir_optim: True
+[2022/03/03 04:21:20] root INFO:  enable_memory_optim: True
+[2022/03/03 04:21:20] root INFO:  enable_tensorrt: False
+[2022/03/03 04:21:20] root INFO:  enable_mkldnn: False
+[2022/03/03 04:21:20] root INFO:  cpu_math_library_num_threads: 1
+[2022/03/03 04:21:20] root INFO: ----------------------- Model info ----------------------
+[2022/03/03 04:21:20] root INFO:  model_name: classification
+[2022/03/03 04:21:20] root INFO:  precision: fp32
+[2022/03/03 04:21:20] root INFO: ----------------------- Data info -----------------------
+[2022/03/03 04:21:20] root INFO:  batch_size: 1
+[2022/03/03 04:21:20] root INFO:  input_shape: dynamic
+[2022/03/03 04:21:20] root INFO:  data_num: 1
+[2022/03/03 04:21:20] root INFO: ----------------------- Perf info -----------------------
+[2022/03/03 04:21:20] root INFO:  cpu_rss(MB): 228.7539, gpu_rss(MB): None, gpu_util: None%
+[2022/03/03 04:21:20] root INFO:  total time spent(s): 0.2199
+[2022/03/03 04:21:20] root INFO:  preprocess_time(ms): 18.5826, inference_time(ms): 201.2458, postprocess_time(ms): 0.0784
+```
+
+该信息可以在运行log中查看，以`mobilenet_v3_small`为例，log位置在`./log/mobilenet_v3_small/lite_train_lite_infer/python_infer_gpu_batchsize_1.log`。
+
 如果运行失败，也会在终端中输出运行失败的日志信息以及对应的运行命令。可以基于该命令，分析运行失败的原因。
+
+`注意`: 混合精度参数配置文件中，默认使用O1模式；O2模式由于部分PR尚未合入，预计PaddlePaddle>=2.2.3版本才会予以接入TIPC测试。

tutorials/mobilenetv3_prod/Step6/test_tipc/test_train_inference_python.sh

@@ -155,7 +155,7 @@ else
         fi
 
         for trainer in ${trainer_list[*]}; do
-            
+        
             run_train=${trainer_py}
             run_export=${norm_export}