Skip to main content

IBM watsonx

watsonx, developed by IBM, offers a variety of pre-trained AI foundation models that can be used for natural language processing (NLP), computer vision, and speech recognition tasks.

Setup​

Accessing watsonx models can be done either through watsonx SaaS on IBM Cloud or using a dedicated watsonx.ai Software instance.

watsonx.ai SaaS - IBM Cloud​

To get started with watsonx SaaS, visit the registration page. If you do not have an existing IBM Cloud account, you can sign up for a free trial.

To authenticate to watsonx.ai SaaS with Continue, you will need to create a project and set up an API key. Then, in continue:

  • Set apiBase to your watsonx SaaS endpoint, e.g. https://us-south.ml.cloud.ibm.com for US South region.
  • Set projectId to your watsonx project ID.
  • Set apiKey to your watsonx API Key.

watsonx.ai Software​

To authenticate to your watsonx.ai Software instance with Continue, you can use either username/password or ZenApiKey method:

  1. Option 1 (Recommended): using ZenApiKey authentication:
    • Set apiBase to your watsonx software endpoint, e.g. https://cpd-watsonx.apps.example.com.
    • Set projectId to your watsonx project ID.
    • Set apiKey to your watsonx Zen API Key. To generate it:
      1. Log in to the CPD web client.
      2. From the toolbar, click your avatar.
      3. Click Profile and settings.
      4. Click API key > Generate new key.
      5. Click Generate.
      6. Click Copy and save your key somewhere safe. You cannot recover this key if you lose it.
      7. Generate your ZenApiKey by running the following command in your preferred terminal: echo "<username>:<apikey>" | base64, replacing <username> with your CPD username and <apikey> with the API Key you just created.
  2. Option 2: using username/password authentication:
    • Set apiBase to your watsonx software endpoint, e.g. https://cpd-watsonx.apps.example.com.
    • Set projectId to your watsonx project ID.
    • Set API Key to your watsonx Username and Password using username:password as format.

Configuration​

Add the following configuration:

Package or config.yaml
models:
  - name: watsonx - Model Name
    provider: watsonx
    model: model ID
    apiBase: https://us-south.ml.cloud.ibm.com
    projectId: PROJECT_ID
    apiKey: API_KEY/ZENAPI_KEY/USERNAME:PASSWORD
    apiVersion: 2024-03-14

apiVersion is optional and defaults to the latest version.

If you are using a custom deployment endpoint, set deploymentID to the model's deployment ID. You can find it in the watsonx.ai Prompt Lab UI by selecting the corresponding model and opening the </> tab on the right, which will display the endpoint's URL containing the deployment ID.

Package or config.yaml
models:
  - name: watsonx - Model Name
    provider: watsonx
    model: model ID
    apiBase: watsonx endpoint e.g. https://us-south.ml.cloud.ibm.com
    apiKey: API_KEY/ZENAPI_KEY/USERNAME:PASSWORD
    apiVersion: 2024-03-14
    deploymentId: DEPLOYMENT_ID

Configuration Options​

Make sure to specify a template name, such as granite or llama3, and to set the contextLength to the model's context window size. You can also configure generation parameters, such as temperature, topP, topK, frequency penalty, and stop sequences:

config.yaml
models:
  - name: Granite Code 20b
    provider: watsonx
    model: ibm/granite-20b-code-instruct
    apiBase: watsonx endpoint e.g. https://us-south.ml.cloud.ibm.com
    projectId: PROJECT_ID
    apiKey: API_KEY/ZENAPI_KEY/USERNAME:PASSWORD
    apiVersion: 2024-03-14
    template: granite
    contextLength: 8000
    defaultCompletionOptions:
      temperature: 0.1
      topP: 0.3
      topK: 20
      maxTokens: 2000
      frequencyPenalty: 1.1
      stop:
        - Question:
        - "\n\n\n"

Tab Auto Complete Model​

Granite models are recommended for tab auto complete. The configuration is similar to that of the chat models:

Package or config.yaml
models:
  - name: Granite Code 8b
    provider: watsonx
    model: ibm/granite-8b-code-instruct
    apiBase: watsonx endpoint e.g. https://us-south.ml.cloud.ibm.com
    projectId: PROJECT_ID
    apiKey: API_KEY/ZENAPI_KEY/USERNAME:PASSWORD
    apiVersion: 2024-03-14
    roles:
      - autocomplete

Embeddings Model​

To view the list of available embeddings models, visit this page.

config.yaml
models:
  - name: Watsonx Embedder
    provider: watsonx
    model: ibm/slate-30m-english-rtrvr-v2
    apiBase: https://us-south.ml.cloud.ibm.com
    projectId: PROJECT_ID
    apiKey: API_KEY/ZENAPI_KEY/USERNAME:PASSWORD
    apiVersion: 2024-03-14
    roles:
      - embed