snowflake.ml.model.custom_model.MethodRef¶

class snowflake.ml.model.custom_model.MethodRef(model_ref: ModelRef, method_name: str)¶

Bases: object

Represents a method invocation of an instance of ModelRef.

This allows us to:

1. Customize the place of actual execution of the method (inline, thread/process pool, or remote).

2. Enrich the way of execution (sync versus async).

Example

If you have an SKL model, you would normally invoke it by skl_ref.predict(df), which has a synchronous API. Within the inference graph, you could invoke await skl_ref.predict.async_run(df), which will automatically run on a thread with an asynchronous interface.

Methods

async async_run(*args: Any, **kwargs: Any) → Any¶

Run the method in an asynchronous way. If the method is defined as async, this will simply run it. If not, this will be run in a separate thread.

Parameters:

• *args – Arguments of the original method.

• **kwargs – Keyword arguments of the original method.

Returns:

The result of the original method.