pipelines/components/aws/sagemaker/train/component.yaml

34 lines
1.8 KiB
YAML

name: 'Sagemaker - Training Job'
description: |
Train Machine Learning and Deep Learning Models using SageMaker
inputs:
- {name: region, description: 'The region where the training job launches.'}
- {name: image, description: 'The registry path of the Docker image that contains the training algorithm.'}
- {name: instance_type, description: 'The ML compute instance type.'}
- {name: instance_count, description: 'The registry path of the Docker image that contains the training algorithm.'}
- {name: volume_size, description: 'The size of the ML storage volume that you want to provision.'}
- {name: dataset_path, description: 'The S3 location of the data source that is associated with a channel.'}
- {name: model_artifact_path, description: 'Identifies the S3 path where you want Amazon SageMaker to store the model artifacts.'}
- {name: role, description: 'The Amazon Resource Name (ARN) that Amazon SageMaker assumes to perform tasks on your behalf.'}
outputs:
- {name: model_artifact_url, description: 'Model artifacts url'}
- {name: job_name, description: 'Training job name'}
implementation:
container:
image: seedjeffwan/kubeflow-pipeline-aws-sm:20190501-05
command: ['python']
args: [
train.py,
--region, {inputValue: region},
--image, {inputValue: image},
--instance_type, {inputValue: instance_type},
--instance_count, {inputValue: instance_count},
--volume_size, {inputValue: volume_size},
--dataset_path, {inputValue: dataset_path},
--model_artifact_path, {inputValue: model_artifact_path},
--role, {inputValue: role}
]
fileOutputs:
model_artifact_url: /tmp/model_artifact_url.txt
job_name: /tmp/job_name.txt