Using ModelArts SDK¶
Notebook instances allow you to use ModelArts SDK to manage OBS, training jobs, models, and real-time services.
Your notebook instances have automatically obtained your AK/SK for authentication and the region. Therefore, SDK sessions are automatically authenticated.
Example Code¶
Create a training job.
from modelarts.session import Session from modelarts.estimator import Estimator session = Session() estimator = Estimator( modelarts_session=session, framework_type='PyTorch', # AI engine name framework_version='PyTorch-1.0.0-python3.6', # AI engine version code_dir='/obs-bucket-name/src/', # Training script directory boot_file='/obs-bucket-name/src/pytorch_sentiment.py', # Training boot script directory log_url='/obs-bucket-name/log/', # Training log directory hyperparameters=[ {"label":"classes", "value": "10"}, {"label":"lr", "value": "0.001"} ], output_path='/obs-bucket-name/output/', # Training output directory train_instance_type='modelarts.vm.gpu.p100', # Training environment specifications train_instance_count=1, # Number of training nodes job_description='pytorch-sentiment with ModelArts SDK') # Training job description job_instance = estimator.fit(inputs='/obs-bucket-name/data/train/', wait=False, job_name='my_training_job')
Obtain a model list.
from modelarts.session import Session from modelarts.model import Model session = Session() model_list_resp = Model.get_model_list(session, model_status="published", model_name="digit", order="desc")
Obtain service details.
from modelarts.session import Session from modelarts.model import Predictor session = Session() predictor_instance = Predictor(session, service_id="input your service_id") predictor_info_resp = predictor_instance.get_service_info()