Deploy Models

Use model deployments when you want to serve a model you own or customize.

Create A Model

from amics import Client

amics = Client()

model = amics.models.deploy(
    "ticket-classifier",
    template="llama",
    region="us-east-1",
)

Create A Deployment

deployment = amics.models.create_deployment(
    model["id"],
    "production",
    accelerator="H100",
    min_replicas=0,
    max_replicas=2,
    production=True,
)

Predict

prediction = amics.models.predict(model["id"], {"input": "Refund request"})

Useful Methods

• amics.models.list()
• amics.models.get(model_id)
• amics.models.prepare_upload(name, filename="model.tar.gz")
• amics.models.deploy(name, template="llama")
• amics.models.create_deployment(model_id, name, accelerator="H100")
• amics.models.predict(model_id, payload)
• amics.models.logs(resource_id)
• amics.models.metrics(resource_id)