diff --git a/test_tipc/supplementary/readme.md b/test_tipc/supplementary/readme.md index b630b0f30b23b71c0dd21def2a45fee01023fe82..cc49f0099685f6de9cf3cf4688c7fbbc6872c429 100644 --- a/test_tipc/supplementary/readme.md +++ b/test_tipc/supplementary/readme.md @@ -47,6 +47,13 @@ bash test_tipc/test_train_python.sh ./test_tipc/train_infer_python_PACT.txt 'lit bash test_tipc/test_train_python.sh ./test_tipc/train_infer_python_FPGM.txt 'lite_train_lite_infer' ``` +多机多卡的运行配置文件分别为'train_infer_python_fleet.txt', 'train_infer_python_FPGM_fleet.txt', 'train_infer_python_PACT_fleet.txt'. +运行时,需要修改配置文件中的`gpu_list:xx.xx.xx.xx,yy.yy.yy.yy;0,1`. 将`xx.xx.xx.xx` 和 `yy.yy.yy.yy`替换为具体的 `ip` 地址。 另外,和单机训练 +不同,多机多卡训练需要在多机的每个节点上分别运行命令。 以多机多卡量化训练为例, 指令如下: +``` +bash test_tipc/test_train_python.sh ./test_tipc/train_infer_python_PACT_fleet.txt 'lite_train_lite_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'模式运行后,在test_tipc/extra_output文件夹有以下文件: ```