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
Studio update launcher tests: give every test a private STUDIO_HOME (#11869)
Daniel Han1ecc01a306a5
diff --git a/tests/python/test_windows_studio_update_launcher.py b/tests/python/test_windows_studio_update_launcher.pyindex 40fd50be15b..b3d6c341505 100644--- a/tests/python/test_windows_studio_update_launcher.py+++ b/tests/python/test_windows_studio_update_launcher.py@@ -55,7 +55,7 @@ REAL_MSVCRT = sys.modules.get("msvcrt")@pytest.fixture-def studio(monkeypatch):+def studio(monkeypatch, tmp_path):package = types.ModuleType("unsloth_cli")package.__path__ = [str(REPO_ROOT / "unsloth_cli")]commands = types.ModuleType("unsloth_cli.commands")@@ -81,6 +81,10 @@ def studio(monkeypatch):module = importlib.util.module_from_spec(spec)monkeypatch.setitem(sys.modules, module_name, module)spec.loader.exec_module(module)+ # update() takes an exclusive flock under STUDIO_HOME, which defaults to the real+ # ~/.unsloth/studio. Another xdist worker holding it made update() exit 1 before+ # the call order these tests check, and the tests wrote into the user's home.+ monkeypatch.setattr(module, "STUDIO_HOME", tmp_path / "studio_home")return module