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 tests: keep two Linux-host assumptions off the Windows and macOS legs (#17107)
Daniel Han8a11e11e49dc
diff --git a/studio/backend/tests/test_diffusion_comfy_block.py b/studio/backend/tests/test_diffusion_comfy_block.pyindex 3bae65e903e..49c08f2ab71 100644--- a/studio/backend/tests/test_diffusion_comfy_block.py+++ b/studio/backend/tests/test_diffusion_comfy_block.py@@ -731,7 +731,11 @@ def _rss_anon_bytes():ctypes.CDLL("libc.so.6").malloc_trim(0) # hand freed heap back, so RssAnon is what is liveexcept OSError:pass- with open("/proc/self/status") as status:+ try:+ status = open("/proc/self/status")+ except OSError: # no procfs (Windows, macOS): callers skip on None+ return None+ with status:for line in status:if line.startswith("RssAnon:"):return int(line.split()[1]) * 1024diff --git a/studio/backend/tests/test_sd_cpp_args.py b/studio/backend/tests/test_sd_cpp_args.pyindex 7431f6dcfd4..ad25c1b09c0 100644--- a/studio/backend/tests/test_sd_cpp_args.py+++ b/studio/backend/tests/test_sd_cpp_args.py@@ -62,6 +62,9 @@ def _physical_metal(monkeypatch):import core.inference.sd_cpp_args as sd_cpp_argsmonkeypatch.setattr(sd_cpp_args, "metal_is_paravirtual", lambda: False)+ # Commands are built for a Linux host unless a test sets the platform, so a Mac runner gets the same+ # commands as Linux (Metal drops direct conv and adds --clip-on-cpu).+ monkeypatch.setattr("sys.platform", "linux")# The real probe is cached and reads sysctl; follow the patched platform.machine() instead.import platform