mlctl 0.0.5

mlctl is the control plane for MLOps. It provides a CLI and a Python SDK for supporting key operations related to MLOps.

mlctl

mlctl is the Command Line Interface (CLI)/Software Development Kit (SDK) for MLOps. It allows for all ML Lifecycle operations, such as Training, Deployment etc. to be controlled via a simple-to-use command line interface. Additionally, mlctl provides a SDK for use in a notebook environment and employs an extensible mechanism for plugging in various back-end providers, such as SageMaker.

The following ML Lifecycle operations are currently supported via mlctl

• train - operations related to model training
• host - operations related to hosting a model for online inference
• batch inference - operations for running model inference in a batch method

Getting Started

Installation

1. (Optional) Create a new virtual environment for mlctl

   pip install virtualenv
   virtualenv ~/envs/mlctl
   source ~/envs/mlctl/bin/activate
   

2. Install mlctl:

   pip install mlctl
   

3. Upgrade an existing version:

   pip install --upgrade mlctl
   

Usage

Optional Setup

mlctl requires users to specify the plugin and a profile/credentials file for authenticating operations. These values can either be stored as environment variables as shown below OR they can be passed as command line options. Use --help for more details.

```
export PLUGIN=
export PROFILE=
```

Commands

mlctl CLI commands have the following structure:

mlctl <command> <subcommand> [OPTIONS]

To view help documentation, run the following:

mlctl --help
mlctl <command> --help
mlctl <command> <subcommand> --help

Initialize ML Model

---

mlctl init [OPTIONS]

Options	Description
template or -t	(optional) Location of the project template github location.

Training Commands

---

mlctl train <subcommand> [OPTIONS]

Subcommand	Description
start	train a model
stop	stop an ongoing training job
info	get training job information

Hosting Commands

---

mlctl hosting <subcommand> [OPTIONS]

Subcommand	Description
create	create a model from trained model artifact
deploy	deploy a model to create an endpoint for inference
undeploy	undeploy a model
info	get endpoint information

Batch Inference Commands

---

mlctl batch <subcommand> [OPTIONS]

Subcommand	Description
start	perform batch inference
stop	stop an ongoing batch inference
info	get batch inference information

Examples

• Sagemaker

Contributing

For information on how to contribute to mlctl, please read through the contributing guidelines.