The ClassificationDeploymentConfig takes in all the same parameters as the DeploymentConfig. The speculator parameter is ignored for classification.
You must use a ClassificationConfig to create a deployment with which classification adapters will be used. You *cannot run classification on a text generation deployment, or run text generation on a classification deployment.

​
Example Usage

from predibase import ClassificationConfig

config = ClassificationConfig(
    base_model="qwen3-8b",
    accelerator="a100_80gb_100",
    min_replicas=0,
    max_replicas=1
)