mirror of
https://github.com/hwchase17/langchain.git
synced 2025-05-01 21:35:34 +00:00
7e7fcf5b1f
- **Description:** Instantiating `GPT4AllEmbeddings` with no `gpt4all_kwargs` argument raised a `ValidationError`. Root cause: #21238 added the capability to pass `gpt4all_kwargs` through to the `GPT4All` instance via `Embed4All`, but broke code that did not specify a `gpt4all_kwargs` argument. - **Issue:** #25119 - **Dependencies:** None - **Twitter handle:** [`@metadaddy`](https://twitter.com/metadaddy)
63 lines
2.0 KiB
Python
63 lines
2.0 KiB
Python
import sys
|
|
from typing import Any, Optional
|
|
from unittest.mock import MagicMock, patch
|
|
|
|
from langchain_community.embeddings import GPT4AllEmbeddings
|
|
|
|
_GPT4ALL_MODEL_NAME = "all-MiniLM-L6-v2.gguf2.f16.gguf"
|
|
_GPT4ALL_NTHREADS = 4
|
|
_GPT4ALL_DEVICE = "gpu"
|
|
_GPT4ALL_KWARGS = {"allow_download": False}
|
|
|
|
|
|
class MockEmbed4All(MagicMock):
|
|
"""Mock Embed4All class."""
|
|
|
|
def __init__(
|
|
self,
|
|
model_name: Optional[str] = None,
|
|
*,
|
|
n_threads: Optional[int] = None,
|
|
device: Optional[str] = None,
|
|
**kwargs: Any,
|
|
): # type: ignore[no-untyped-def]
|
|
assert model_name == _GPT4ALL_MODEL_NAME
|
|
|
|
|
|
class MockGpt4AllPackage(MagicMock):
|
|
"""Mock gpt4all package."""
|
|
|
|
Embed4All = MockEmbed4All
|
|
|
|
|
|
def test_create_gpt4all_embeddings_no_kwargs() -> None:
|
|
"""Test fix for #25119"""
|
|
with patch.dict(sys.modules, {"gpt4all": MockGpt4AllPackage()}):
|
|
embedding = GPT4AllEmbeddings( # type: ignore[call-arg]
|
|
model_name=_GPT4ALL_MODEL_NAME,
|
|
n_threads=_GPT4ALL_NTHREADS,
|
|
device=_GPT4ALL_DEVICE,
|
|
)
|
|
|
|
assert embedding.model_name == _GPT4ALL_MODEL_NAME
|
|
assert embedding.n_threads == _GPT4ALL_NTHREADS
|
|
assert embedding.device == _GPT4ALL_DEVICE
|
|
assert embedding.gpt4all_kwargs == {}
|
|
assert isinstance(embedding.client, MockEmbed4All)
|
|
|
|
|
|
def test_create_gpt4all_embeddings_with_kwargs() -> None:
|
|
with patch.dict(sys.modules, {"gpt4all": MockGpt4AllPackage()}):
|
|
embedding = GPT4AllEmbeddings( # type: ignore[call-arg]
|
|
model_name=_GPT4ALL_MODEL_NAME,
|
|
n_threads=_GPT4ALL_NTHREADS,
|
|
device=_GPT4ALL_DEVICE,
|
|
gpt4all_kwargs=_GPT4ALL_KWARGS,
|
|
)
|
|
|
|
assert embedding.model_name == _GPT4ALL_MODEL_NAME
|
|
assert embedding.n_threads == _GPT4ALL_NTHREADS
|
|
assert embedding.device == _GPT4ALL_DEVICE
|
|
assert embedding.gpt4all_kwargs == _GPT4ALL_KWARGS
|
|
assert isinstance(embedding.client, MockEmbed4All)
|