Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.
Python agentaichatgptdeepseekfine-tuninggemmaimage-generationllamallmllmsopenaipythonqwenreinforcement-learningself-hostedstable-diffusiontext-to-speechttsuiunsloth
# SPDX-License-Identifier: AGPL-3.0-only
# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
from __future__ import annotations
import base64
import io
import pytest
from core.inference.diffusion import decode_b64_image
from core.inference.diffusion_controlnet import preprocess_control
PIL = pytest.importorskip("PIL.Image")
np = pytest.importorskip("numpy")
def _data_url(img) -> str:
buf = io.BytesIO()
img.save(buf, format = "PNG")
return "data:image/png;base64," + base64.b64encode(buf.getvalue()).decode()
def _black_ink_on_transparent():
img = PIL.new("RGBA", (200, 80), (0, 0, 0, 0))
img.paste((0, 0, 0, 255), (60, 20, 140, 60))
return img
def test_a_transparent_source_reaches_the_model_on_white():
img = decode_b64_image(_data_url(_black_ink_on_transparent()))
assert img.mode == "RGB"
assert img.getpixel((5, 5)) == (255, 255, 255)
assert img.getpixel((100, 40)) == (0, 0, 0)
def test_transparent_line_art_still_gives_a_canny_edge_map():
sketch = PIL.new("RGBA", (128, 128), (0, 0, 0, 0))
for x in range(20, 108):
for y in (30, 31, 96, 97):
sketch.putpixel((x, y), (0, 0, 0, 255))
edges = preprocess_control(decode_b64_image(_data_url(sketch)), "canny")
assert np.asarray(edges).max() == 255
def test_a_paletted_transparent_png_reaches_the_model_on_white():
img = decode_b64_image(_data_url(_black_ink_on_transparent().quantize(colors = 4)))
assert img.getpixel((5, 5)) == (255, 255, 255)
assert img.getpixel((100, 40)) == (0, 0, 0)
def test_rgba_and_mask_decodes_keep_their_alpha_and_values():
data = _data_url(_black_ink_on_transparent())
assert decode_b64_image(data, mode = "RGBA").getpixel((5, 5)) == (0, 0, 0, 0)
assert decode_b64_image(data, mode = "L").getpixel((5, 5)) == 0
def test_an_opaque_rgba_canvas_export_decodes_like_rgb():
img = PIL.new("RGBA", (64, 48), (10, 200, 30, 255))
img.paste((200, 0, 120, 255), (16, 12, 48, 36))
assert decode_b64_image(_data_url(img)).tobytes() == img.convert("RGB").tobytes()