Use AI Labs as a learner

Work in the courses an Admin adds you to.

Before you begin

Make sure that an Admin has added you to an AI Labs organization and to the course you want to work in.

View your courses and course material

Open your courses to read the material your Professor shared and download the files you need.

• Accept the invitation in the email you received. You see the organizations you belong to, and in each one only the courses you were added to.
• Sign in to AI Labs, which is available in Delhi.

Check your dashboard

Your Dashboard gives an overview of your active resources, notifications, and courses:

• Services Running has your nodes, inferences, and datasets, with how many are running and the total.
• Courses Overview has the courses you were added to, each with its instructor and a short description. Select a course to open it.

Open a course

1. In the left menu, select Courses.
2. Select a course. Its Courseware page lists the modules and study material.
3. Open a module and download its files, such as lecture PDFs, slides, or notes.

You can view and download course material. Your Admin or Professor uploads and changes it.

For example, in the course Intro to Cloud Computing, open Module 1 and download materials such as Module 1/06.pdf.

Practise what you learn: launch an instance. If a course is missing, see Troubleshooting AI Labs.

Launch an instance

Launch an instance to get a ready-to-use workspace for AI and machine learning work, with containers, JupyterLab, and popular AI frameworks included.

• Make sure that your group gives you Create access to Node, the service for instances.
• Check that the organization is running. You can create and start instances only while it runs.
• Check that you are within your instance limit for the plan you want. Your group sets how many instances you can run on each plan.
• Create or choose a dataset if you want one on the instance. See Create a dataset.

Launch the instance

1. In the left menu, select Instances, then Create Instance.
2. Choose an image. AI Labs instances use ready-made images.
3. Choose a plan with the GPU memory, RAM, and CPU you need, from the plans your groups allow. To narrow the choice, filter by CPU, RAM, or GPU memory.
4. Choose the workspace size, and choose an existing dataset or create a new one if you want it.
5. Enter a name for the instance and continue.
6. Review the summary and select Launch.

The instance appears in your Instances list and starts.

If the instance does not launch

• You have reached your allowed instance quota and cannot launch any additional resources at this time. means you have used the limit your group gives you on this plan. Delete an instance you are finished with, or ask an Admin to raise the limit.
• This AI Labs organization is currently stopped, so you cannot start this resource. means an Admin stopped the organization. Ask the Admin to start it.

For more, see Troubleshooting AI Labs.

Add data with Create a dataset or deploy a model with Create a model repository.

Create a dataset

Create a dataset to keep your data in one place and open it from your instances and training code.

• Make sure that your group gives you Create access to Storage.
• Decide the storage type. A dataset is backed by object storage or by disk storage.

You can create datasets even when an Admin has stopped the organization. Datasets that you create are yours, and they stay hidden from other learners.

Create the dataset

1. In the left menu, select Dataset, then Launch Dataset.
2. Enter a dataset name and choose the storage type.
3. Select Create.

The dataset appears in your list and is ready to mount on an instance.

Mount the dataset when you launch an instance. For messages, see Troubleshooting AI Labs.

Create a model repository

Create a model repository to store model weights and configuration files, so an endpoint can serve the model.

Make sure that your group gives you Create access to Inference.

Create the repository

1. In the left menu, select Inference, then Model Repository.
2. Select Create Repository.
3. Enter a repository name, choose the model type, and choose the storage type: object storage or disk storage. If you choose an existing object storage bucket, AI Labs shows Invalid storage type.
4. Select Create.

Upload your model files, then create a model endpoint that serves them. For messages, see Troubleshooting AI Labs.

Create a model endpoint

Create a model endpoint to give your applications an API that sends data to your deployed model and gets inference results back in real time.

• Make sure that your group gives you Create access to Inference.
• Create a model repository that holds the model.
• Check that the organization is running and that you are within your instance limit for the plan you choose. An endpoint counts each replica against your limit.

Create the endpoint

1. In the left menu, select Inference, then Model Endpoints.
2. Select Create Endpoint.
3. Choose the framework.
4. Select Link with Model Repository and choose your repository.
5. Choose a plan and continue.
6. (Optional) Set the LLM settings for a language model.
7. Set Active Workers and Max Workers.
8. (Optional) Add environment variables, then continue.
9. Check the summary. To change a step, select its edit button.
10. Select Launch.

An endpoint in AI Labs cannot use a shared file system or a private container.

When the replicas exceed your limit, AI Labs shows You cannot select more than … replicas as you have reached your allowed instance quota. Choose a smaller number.

Try the model with the Playground. For messages, see Troubleshooting AI Labs.

Chat with a model in the Playground

Chat with a deployed chat model in the Playground to try it before you connect an application to it.

Create a model endpoint for a chat model and wait for it to run.

Chat with the model

1. In the left menu, select Inference, then Playground.
2. Select a model.
3. Set the parameters you want to use.
4. Start generating responses.

To change how the model runs, see Create a model endpoint. If a message appears, see Troubleshooting AI Labs.

Create an integration

Create an integration to connect an external platform such as Hugging Face or Weights & Biases. Use it for model training, tracking, and deployment on your AI Labs instances.

• Make sure that your group gives you Create access to External Integration.
• Have the token for the platform, for example a Hugging Face token.

Create the integration

1. In the left menu, select Integrations, then Create Integration.
2. Choose the Integration Type from the list.
3. Enter your token, for example the Hugging Face token.
4. Select Create.

The integration is ready to use in your AI Labs work. If a message appears, see Troubleshooting AI Labs.

Create a personal access token

Create a personal access token from the Personal Access Tokens page.

Make sure that your group gives you Create access to API Token.

Create the token

1. In the left menu, select Personal Access Tokens, then Create Token.
2. Enter a token name.
3. Select Create.

The token appears in your list. Only you see the tokens you create.

Keep the token private. If a message appears, see Troubleshooting AI Labs.