langchain/libs/community/tests/unit_tests/llms/test_databricks.py
Liang Zhang 81985b31e6
community[patch]: Databricks SerDe uses cloudpickle instead of pickle (#18607)
- **Description:** Databricks SerDe uses cloudpickle instead of pickle
when serializing a user-defined function transform_input_fn since pickle
does not support functions defined in `__main__`, and cloudpickle
supports this.
- **Dependencies:** cloudpickle>=2.0.0

Added a unit test.
2024-03-05 18:04:45 -08:00

57 lines
1.6 KiB
Python

"""test Databricks LLM"""
from typing import Any, Dict
import pytest
from pytest import MonkeyPatch
from langchain_community.llms.databricks import (
Databricks,
_load_pickled_fn_from_hex_string,
)
class MockDatabricksServingEndpointClient:
def __init__(
self,
host: str,
api_token: str,
endpoint_name: str,
databricks_uri: str,
task: str,
):
self.host = host
self.api_token = api_token
self.endpoint_name = endpoint_name
self.databricks_uri = databricks_uri
self.task = task
def transform_input(**request: Any) -> Dict[str, Any]:
request["messages"] = [{"role": "user", "content": request["prompt"]}]
del request["prompt"]
return request
@pytest.mark.requires("cloudpickle")
def test_serde_transform_input_fn(monkeypatch: MonkeyPatch) -> None:
import cloudpickle
monkeypatch.setattr(
"langchain_community.llms.databricks._DatabricksServingEndpointClient",
MockDatabricksServingEndpointClient,
)
monkeypatch.setenv("DATABRICKS_HOST", "my-default-host")
monkeypatch.setenv("DATABRICKS_TOKEN", "my-default-token")
llm = Databricks(
endpoint_name="databricks-mixtral-8x7b-instruct",
transform_input_fn=transform_input,
)
params = llm._default_params
pickled_string = cloudpickle.dumps(transform_input).hex()
assert params["transform_input_fn"] == pickled_string
request = {"prompt": "What is the meaning of life?"}
fn = _load_pickled_fn_from_hex_string(params["transform_input_fn"])
assert fn(**request) == transform_input(**request)