示例代码

在ModelArts notebook平台，Session鉴权无需输入鉴权参数。其它平台的Session鉴权请参见Session鉴权。

• 示例一：使用OBS存储位置创建训练作业版本

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

  from modelarts.session import Session from modelarts.estimator import Estimator session = Session() estimator = Estimator( modelarts_session=session, framework_type='PyTorch', # AI引擎名称 framework_version='PyTorch-1.0.0-python3.6', # AI引擎版本 code_dir='/bucket/src/', # 训练脚本目录 boot_file='/bucket/src/pytorch_sentiment.py', # 训练启动脚本目录 log_url='/bucket/log/', # 训练日志目录 hyperparameters=[ {"label":"classes", "value": "10"}, {"label":"lr", "value": "0.001"} ], output_path='/bucket/output/', # 训练输出目录 train_instance_type='modelarts.vm.gpu.p100', # 训练环境规格 train_instance_count=1) job_version_instance = estimator.create_job_version(job_id='182626', pre_version_id=278813, inputs='/bucket/data/train/', wait=False, job_desc='create a job version')

• 示例二：使用数据集创建训练作业版本

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

  from modelarts.session import Session from modelarts.estimator import Estimator session = Session() estimator = Estimator( modelarts_session=session, framework_type='PyTorch', # AI引擎名称 framework_version='PyTorch-1.0.0-python3.6', # AI引擎版本 code_dir='/bucket/src/', # 训练脚本目录 boot_file='/bucket/src/pytorch_sentiment.py', # 训练启动脚本目录 log_url='/bucket/log/', # 训练日志目录 hyperparameters=[ {"label":"classes", "value": "10"}, {"label":"lr", "value": "0.001"} ], output_path='/bucket/output/', # 训练输出目录 train_instance_type='modelarts.vm.gpu.p100', # 训练环境规格 train_instance_count=1, # 训练节点个数 job_description='pytorch-sentiment with ModelArts SDK') # 训练作业描述 job_version_instance = estimator.create_job_version(job_id='182626', pre_version_id=278813, inputs='/bucket/data/train/', wait=False, job_desc='create a job version')

• 示例三：创建 自定义镜像 的训练作业版本

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

  from modelarts.session import Session from modelarts.estimator import Estimator session = Session() estimator = Estimator( modelarts_session=session, log_url='/bucket/log/', # 训练日志目录 hyperparameters=[ {"label":"classes", "value": "10"}, {"label":"lr", "value": "0.001"} ], output_path='/bucket/output/', # 训练输出目录 train_instance_type='modelarts.vm.gpu.p100', # 训练环境规格 train_instance_count=1, # 训练节点个数 user_command='bash -x /home/work/run_train.sh python /home/work/user-job-dir/app/mnist/mnist_softmax.py --data_url /home/work/user-job-dir/app/mnist_data', # 自定义镜像启动命令 user_image_url='100.125.5.235:20202/jobmng/cpu-base:1.0', # 自定义镜像下载地址 job_description='pytorch-sentiment with ModelArts SDK') # 训练作业描述 job_version_instance = estimator.create_job_version(job_id='182626', pre_version_id=278813, inputs='/bucket/data/train/', wait=False, job_desc='create a job version')

• 示例四：使用内置模型创建训练作业版本

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

  from modelarts.session import Session from modelarts.estimator import Estimator session = Session() estimator = Estimator( modelarts_session=session, model_name='Faster_RCNN_ResNet_v1_50', # 内置模型名称 log_url='/bucket/log/', # 训练日志目录 hyperparameters=[ {"label":"classes", "value": "10"}, {"label":"lr", "value": "0.001"} ], output_path='/bucket/output/', # 训练输出目录 train_instance_type='modelarts.vm.gpu.p100', # 训练环境规格 train_instance_count=1, # 训练节点个数 job_description='pytorch-sentiment with ModelArts SDK') # 训练作业描述 job_version_instance = estimator.create_job_version(job_id='182626', pre_version_id=278813, inputs='/bucket/data/train/', wait=False, job_desc='create a job version')