mirror of
https://github.com/invoke-ai/InvokeAI
synced 2024-08-30 20:32:17 +00:00
45 lines
1.7 KiB
Python
45 lines
1.7 KiB
Python
from pathlib import Path
|
|
|
|
import pytest
|
|
|
|
from invokeai.backend.model_manager import BaseModelType, ModelRepoVariant
|
|
from invokeai.backend.model_manager.probe import ControlAdapterProbe, VaeFolderProbe
|
|
|
|
|
|
@pytest.mark.parametrize(
|
|
"vae_path,expected_type",
|
|
[
|
|
("sd-vae-ft-mse", BaseModelType.StableDiffusion1),
|
|
("sdxl-vae", BaseModelType.StableDiffusionXL),
|
|
("taesd", BaseModelType.StableDiffusion1),
|
|
("taesdxl", BaseModelType.StableDiffusionXL),
|
|
],
|
|
)
|
|
def test_get_base_type(vae_path: str, expected_type: BaseModelType, datadir: Path):
|
|
sd1_vae_path = datadir / "vae" / vae_path
|
|
probe = VaeFolderProbe(sd1_vae_path)
|
|
base_type = probe.get_base_type()
|
|
assert base_type == expected_type
|
|
repo_variant = probe.get_repo_variant()
|
|
assert repo_variant == ModelRepoVariant.Default
|
|
|
|
|
|
def test_repo_variant(datadir: Path):
|
|
probe = VaeFolderProbe(datadir / "vae" / "taesdxl-fp16")
|
|
repo_variant = probe.get_repo_variant()
|
|
assert repo_variant == ModelRepoVariant.FP16
|
|
|
|
|
|
def test_controlnet_t2i_default_settings():
|
|
should_be_canny = ControlAdapterProbe.get_default_settings("some_canny_model")
|
|
assert should_be_canny and should_be_canny.preprocessor == "canny_image_processor"
|
|
|
|
should_be_depth_anything = ControlAdapterProbe.get_default_settings("some_depth_model")
|
|
assert should_be_depth_anything and should_be_depth_anything.preprocessor == "depth_anything_image_processor"
|
|
|
|
should_be_dw_openpose = ControlAdapterProbe.get_default_settings("some_pose_model")
|
|
assert should_be_dw_openpose and should_be_dw_openpose.preprocessor == "dw_openpose_image_processor"
|
|
|
|
should_be_none = ControlAdapterProbe.get_default_settings("i like turtles")
|
|
assert should_be_none is None
|