149 lines
6.4 KiB
YAML
149 lines
6.4 KiB
YAML
name: SageMaker - RLEstimator Training Job
|
|
description: Handle end-to-end training and deployment of custom RLEstimator code.
|
|
inputs:
|
|
- name: spot_instance
|
|
type: Bool
|
|
description: Use managed spot training.
|
|
default: "False"
|
|
- {name: max_wait_time, type: Integer, description: The maximum time in seconds you
|
|
are willing to wait for a managed spot training job to complete., default: '86400'}
|
|
- {name: max_run_time, type: Integer, description: The maximum run time in seconds
|
|
for the training job., default: '86400'}
|
|
- {name: checkpoint_config, type: JsonObject, description: Dictionary of information
|
|
about the output location for managed spot training checkpoint data., default: '{}'}
|
|
- {name: region, type: String, description: The region for the SageMaker resource.}
|
|
- {name: endpoint_url, type: String, description: The URL to use when communicating
|
|
with the SageMaker service., default: ''}
|
|
- {name: assume_role, type: String, description: The ARN of an IAM role to assume
|
|
when connecting to SageMaker., default: ''}
|
|
- {name: tags, type: JsonObject, description: 'An array of key-value pairs, to categorize
|
|
AWS resources.', default: '{}'}
|
|
- {name: job_name, type: String, description: Training job name., default: ''}
|
|
- {name: role, type: String, description: The Amazon Resource Name (ARN) that Amazon
|
|
SageMaker assumes to perform tasks on your behalf.}
|
|
- {name: image, type: String, description: 'An ECR url. If specified, the estimator
|
|
will use this image for training and hosting', default: ''}
|
|
- {name: entry_point, type: String, description: Path (absolute or relative) to the
|
|
Python source file which should be executed as the entry point to training., default: ''}
|
|
- {name: source_dir, type: String, description: Path (S3 URI) to a directory with
|
|
any other training source code dependencies aside from the entry point file.,
|
|
default: ''}
|
|
- {name: toolkit, type: String, description: RL toolkit you want to use for executing
|
|
your model training code., default: ''}
|
|
- {name: toolkit_version, type: String, description: RL toolkit version you want to
|
|
be use for executing your model training code., default: ''}
|
|
- {name: framework, type: String, description: 'Framework (MXNet, TensorFlow or PyTorch)
|
|
you want to be used as a toolkit backed for reinforcement learning training.',
|
|
default: ''}
|
|
- {name: metric_definitions, type: JsonArray, description: The dictionary of name-regex
|
|
pairs specify the metrics that the algorithm emits., default: '[]'}
|
|
- {name: training_input_mode, type: String, description: The input mode that the algorithm
|
|
supports. File or Pipe., default: File}
|
|
- {name: hyperparameters, type: JsonObject, description: Hyperparameters that will
|
|
be used for training., default: '{}'}
|
|
- {name: instance_type, type: String, description: The ML compute instance type.,
|
|
default: ml.m4.xlarge}
|
|
- {name: instance_count, type: Integer, description: The number of ML compute instances
|
|
to use in the training job., default: '1'}
|
|
- {name: volume_size, type: Integer, description: The size of the ML storage volume
|
|
that you want to provision., default: '30'}
|
|
- {name: max_run, type: Integer, description: 'Timeout in seconds for training (default:
|
|
24 * 60 * 60).', default: '86400'}
|
|
- {name: model_artifact_path, type: String, description: Identifies the S3 path where
|
|
you want Amazon SageMaker to store the model artifacts.}
|
|
- {name: vpc_security_group_ids, type: JsonArray, description: 'The VPC security group
|
|
IDs, in the form sg-xxxxxxxx.', default: '[]'}
|
|
- {name: vpc_subnets, type: JsonArray, description: The ID of the subnets in the VPC
|
|
to which you want to connect your hpo job., default: '[]'}
|
|
- name: network_isolation
|
|
type: Bool
|
|
description: Isolates the training container.
|
|
default: "False"
|
|
- name: traffic_encryption
|
|
type: Bool
|
|
description: Encrypts all communications between ML compute instances in distributed
|
|
training.
|
|
default: "False"
|
|
- {name: debug_hook_config, type: JsonObject, description: 'Configuration information
|
|
for the debug hook parameters, collection configuration, and storage paths.',
|
|
default: '{}'}
|
|
- {name: debug_rule_config, type: JsonArray, description: Configuration information
|
|
for debugging rules., default: '[]'}
|
|
outputs:
|
|
- {name: model_artifact_url, description: The model artifacts URL.}
|
|
- {name: job_name, description: The training job name.}
|
|
- {name: training_image, description: The registry path of the Docker image that contains
|
|
the training algorithm.}
|
|
implementation:
|
|
container:
|
|
image: public.ecr.aws/kubeflow-on-aws/aws-sagemaker-kfp-components:1.1.2
|
|
command: [python3]
|
|
args:
|
|
- rlestimator/src/sagemaker_rlestimator_component.py
|
|
- --spot_instance
|
|
- {inputValue: spot_instance}
|
|
- --max_wait_time
|
|
- {inputValue: max_wait_time}
|
|
- --max_run_time
|
|
- {inputValue: max_run_time}
|
|
- --checkpoint_config
|
|
- {inputValue: checkpoint_config}
|
|
- --region
|
|
- {inputValue: region}
|
|
- --endpoint_url
|
|
- {inputValue: endpoint_url}
|
|
- --assume_role
|
|
- {inputValue: assume_role}
|
|
- --tags
|
|
- {inputValue: tags}
|
|
- --job_name
|
|
- {inputValue: job_name}
|
|
- --role
|
|
- {inputValue: role}
|
|
- --image
|
|
- {inputValue: image}
|
|
- --entry_point
|
|
- {inputValue: entry_point}
|
|
- --source_dir
|
|
- {inputValue: source_dir}
|
|
- --toolkit
|
|
- {inputValue: toolkit}
|
|
- --toolkit_version
|
|
- {inputValue: toolkit_version}
|
|
- --framework
|
|
- {inputValue: framework}
|
|
- --metric_definitions
|
|
- {inputValue: metric_definitions}
|
|
- --training_input_mode
|
|
- {inputValue: training_input_mode}
|
|
- --hyperparameters
|
|
- {inputValue: hyperparameters}
|
|
- --instance_type
|
|
- {inputValue: instance_type}
|
|
- --instance_count
|
|
- {inputValue: instance_count}
|
|
- --volume_size
|
|
- {inputValue: volume_size}
|
|
- --max_run
|
|
- {inputValue: max_run}
|
|
- --model_artifact_path
|
|
- {inputValue: model_artifact_path}
|
|
- --vpc_security_group_ids
|
|
- {inputValue: vpc_security_group_ids}
|
|
- --vpc_subnets
|
|
- {inputValue: vpc_subnets}
|
|
- --network_isolation
|
|
- {inputValue: network_isolation}
|
|
- --traffic_encryption
|
|
- {inputValue: traffic_encryption}
|
|
- --debug_hook_config
|
|
- {inputValue: debug_hook_config}
|
|
- --debug_rule_config
|
|
- {inputValue: debug_rule_config}
|
|
- --model_artifact_url_output_path
|
|
- {outputPath: model_artifact_url}
|
|
- --job_name_output_path
|
|
- {outputPath: job_name}
|
|
- --training_image_output_path
|
|
- {outputPath: training_image}
|