SkyPilot¶
SkyPilot is a framework tailored for executing Large Language Models (LLMs) across various cloud platforms, ensuring optimal cost efficiency, abundant GPU availability, and streamlined execution management. It assists in provisioning tasks to utilize the most advantageous resources available while minimizing expenses.
Prerequisites¶
You must have a SkyPilot-configurable cloud account.
Note
Currently SkyPilot supports the following cloud infrastructure: Kubernetes; AWS, GCP, Azure, OCI, Lambda Cloud, Fluidstack, RunPod, Cudo, Paperspace, Cloudflare, Samsung, IBM, VMware vSphere. For more information, visit the SkyPilot documentation
If you are in an organization, an IAM User Account is needed to set up SkyPilot policy for the account.
For example; If you have an AWS IAM user account, use the following URL:
Also, in case of using AWS cloud account, an Access Key and Secret Access Key is required to access the AWS account resources.
To obtain the access key and secret access key, click on the following page and follow the steps provided:
Configuring SkyPilot on DKubeX¶
To set up SkyPilot on your DKubeX setup, use the following steps.
On your DKubeX backend, execute into the
sky-controller-v2-0
pod.kubectl exec -it -n d3x sky-controller-v2-0 -- bash
Use the following command and provide your AWS access key and secret key. Rest of the fields can be left blank as default.
aws configure
Workflow Tutorials with SkyPilot¶
Visit the links provided below for detailed tutorials regarding running specific workflows in DKubeX with SkyPilot.
Tutorial on how to run ingestion jobs and create datasets on SkyPilot clusters.
Tutorial on deploying embedding models on DKubeX with SkyPilot and Sky-Serve.
Tutorial on how to deploy LLMs on DKubeX with SkyPilot and Sky-Serve.
Tutorial on running Sky jobs regarding finetuning Wine and Llama-2 models on DKubeX.