About Running Jobs
You run jobs (experiments) with MissingLink in one of two modes:
On a local workstation: Suppose you are developing and debugging your model, you require limited resources and wish to maintain tight control over run parameters and data. During this phase of your experimentation, a local workstation might suit your needs.
Even when working on a local machine, running the experiment through MissingLink will enable you to manage the experiment artifacts and document all the runtime parameters to fully support your experiment analysis and planning.
For running jobs locally, use the
ml run localcommand. For more information, see Running Jobs on a Workstation.
Once you have successfully run the experiment on your workstation, you are ready to run experiments in the cloud.
In the cloud: "Cloud" can mean either a local grid of on-premises servers, call it a 'private cloud'; or a cloud service such as AWS, Google, and Azure. Utilizing the cloud enables you to run multiple experiments in parallel, without having to know where the machine is, set it up, or even connect to it.
To run jobs in the cloud, you must:
- Perform cloud authorization. For more information, see Creating Private Cloud Authorization Using On-premises Machines and Creating Cloud Authorization for AWS.
- Use the
ml run xpcommand. For more information, see Running Jobs in the Cloud.