Deploy Inference Endpoint and Wait
A 2-step recipe with Hugging Face. We can help you get access.
Recipe · 2 steps
Deploy Inference Endpoint and Wait
with Hugging Face · Artificial Intelligence Management
Create a dedicated Inference Endpoint and wait until it is running.
- Create a new dedicated Inference Endpoint for the model. A 201 returns the endpoint record; a 409 means an endpoint with this name already exists.
- Poll the endpoint status. While the state is pending or initializing the step loops back to itself; once the state is running it proceeds to the end of the workflow.
What you'll need to have ready
- hfToken: Hugging Face access token used as a Bearer credential.
- namespace: User or organization namespace that will own the endpoint.
- name: The endpoint name to create.
- repository: The model repository to deploy (e.g. gpt2).
- revision: Git revision of the model to deploy.
- task: The pipeline task the endpoint will serve.
- vendor: Cloud vendor to deploy on.
- region: Cloud region to deploy in.
- instanceType: Compute instance type identifier.
- instanceSize: Compute instance size identifier.
For your developer
Arazzo workflow hugging-face:deploy-inference-endpoint, 2 operations: createEndpoint → getEndpoint. Specs on APIs.io.