Linux端基础训练预测功能测试的主程序为test_train_inference_python.sh
,可以测试基于Python的模型训练、评估、推理等基本功能。
运行环境配置请参考文档的内容配置TIPC的运行环境。
- 安装PaddlePaddle >= 2.0
- 安装autolog(规范化日志输出工具)
git clone https://github.com/LDOUBLEV/AutoLog cd AutoLog pip3 install -r requirements.txt python3 setup.py bdist_wheel pip3 install ./dist/auto_log-1.0.0-py3-none-any.whl cd ../
先运行prepare.sh
准备数据和模型,然后运行test_train_inference_python.sh
进行测试,最终在test_tipc/output
目录下生成python_infer_*.log
格式的日志文件。
test_train_inference_python.sh
包含5种运行模式,每种模式的运行数据不同,分别用于测试速度和精度,分别是:
- 模式1:lite_train_lite_infer,使用少量数据训练,用于快速验证训练到预测的走通流程,不验证精度和速度;
bash test_tipc/prepare.sh ./test_tipc/configs/wide_deep/train_infer_python.txt 'lite_train_lite_infer'
bash test_tipc/test_train_inference_python.sh ./test_tipc/configs/wide_deep/train_infer_python.txt 'lite_train_lite_infer'
- 模式2:lite_train_whole_infer,使用少量数据训练,一定量数据预测,用于验证训练后的模型执行预测,预测速度是否合理;
bash test_tipc/prepare.sh ./test_tipc/configs/wide_deep/train_infer_python.txt 'lite_train_whole_infer'
bash test_tipc/test_train_inference_python.sh ./test_tipc/configs/wide_deep/train_infer_python.txt 'lite_train_whole_infer'
- 模式3:whole_infer,不训练,全量数据预测,走通开源模型评估、动转静,检查inference model预测时间和精度;
bash test_tipc/prepare.sh ./test_tipc/configs/wide_deep/train_infer_python.txt 'whole_infer'
# 用法1:
bash test_tipc/test_train_inference_python.sh ../test_tipc/configs/wide_deep/train_infer_python.txt 'whole_infer'
# 用法2: 指定GPU卡预测,第三个传入参数为GPU卡号
bash test_tipc/test_train_inference_python.sh ./test_tipc/configs/wide_deep/train_infer_python.txt 'whole_infer' '1'
- 模式4:whole_train_whole_infer,CE: 全量数据训练,全量数据预测,验证模型训练精度,预测精度,预测速度;
bash test_tipc/prepare.sh ./test_tipc/configs/wide_deep/train_infer_python.txt 'whole_train_whole_infer'
bash test_tipc/test_train_inference_python.sh ./test_tipc/configs/wide_deep/train_infer_python.txt 'whole_train_whole_infer'
运行相应指令后,在test_tipc/output
文件夹下自动会保存运行日志。如'lite_train_lite_infer'模式下,会运行训练+inference的链条,因此,在test_tipc/output
文件夹有以下文件:
test_tipc/output/
|- results_python.log # 运行指令状态的日志
|- norm_train_gpus_0_autocast_null/ # GPU 0号卡上正常训练的训练日志和模型保存文件夹
|- pact_train_gpus_0_autocast_null/ # GPU 0号卡上量化训练的训练日志和模型保存文件夹
......
|- python_infer_cpu_usemkldnn_True_threads_1_batchsize_1.log # CPU上开启Mkldnn线程数设置为1,测试batch_size=1条件下的预测运行日志
|- python_infer_gpu_usetrt_True_precision_fp16_batchsize_1.log # GPU上开启TensorRT,测试batch_size=1的半精度预测日志
......
其中results_python.log
中包含了每条指令的运行状态,如果运行成功会输出:
Run successfully with command - python3.7 -u tools/trainer.py -m ./models/rank/wide_deep/config_bigdata.yaml -o runner.print_interval=2 runner.use_gpu=True runner.model_save_path=./test_tipc/output/norm_train_gpus_0_autocast_False runner.epochs=4 auto_cast=False runner.train_batch_size=50 runner.train_data_dir=../../../test_tipc/data/train
Run successfully with command - python3.7 -u tools/to_static.py -m ./models/rank/wide_deep/config_bigdata.yaml -o runner.CE=true runner.model_init_path=./test_tipc/output/norm_train_gpus_0_autocast_False/3 runner.model_save_path=./test_tipc/output/norm_train_gpus_0_autocast_False
......
如果运行失败,会输出:
Run failed with command - python3.7 -u tools/trainer.py -m ./models/rank/wide_deep/config_bigdata.yaml -o runner.print_interval=2 runner.use_gpu=True runner.model_save_path=./test_tipc/output/norm_train_gpus_0_autocast_False runner.epochs=4 auto_cast=False runner.train_batch_size=50 runner.train_data_dir=../../../test_tipc/data/train
Run failed with command - python3.7 -u tools/to_static.py -m ./models/rank/wide_deep/config_bigdata.yaml -o runner.CE=true runner.model_init_path=./test_tipc/output/norm_train_gpus_0_autocast_False/3 runner.model_save_path=./test_tipc/output/norm_train_gpus_0_autocast_False
......
可以很方便的根据results_python.log
中的内容判定哪一个指令运行错误。
本文档为功能测试用,更丰富的训练预测使用教程请参考:
模型训练
基于Python预测引擎推理