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
Give the launcher signal tests a worker to themselves (#11085)
Daniel Han9d72fadf4bc0
diff --git a/.github/workflows/studio-backend-ci.yml b/.github/workflows/studio-backend-ci.ymlindex 59ebedb5ca3..0dfed4699bc 100644--- a/.github/workflows/studio-backend-ci.yml+++ b/.github/workflows/studio-backend-ci.yml@@ -373,7 +373,7 @@ jobs:# yet. It only ever gets looser as the work per shard drops.run: |timeout --signal=INT --kill-after=60 1500 \- python -m pytest ${{ matrix.selection }} -q --tb=short -n 4 --timeout=330 \+ python -m pytest ${{ matrix.selection }} -q --tb=short -n 4 --dist loadgroup --timeout=330 \--ignore=tests/test_studio_api.py \--ignore=tests/test_streaming_stripper.py \--ignore=tests/test_llama_cpp_wait_for_vram_settle.py \@@ -655,7 +655,7 @@ jobs:# under the job cap, so this fires first and says where it was stuck.run: |timeout --signal=INT --kill-after=60 2100 \- python -m pytest ${{ matrix.selection }} -q --tb=short -n 4 --timeout=330 \+ python -m pytest ${{ matrix.selection }} -q --tb=short -n 4 --dist loadgroup --timeout=330 \-m 'not server and not e2e'- name: Hardware-spoof tests (state-sensitive, run in isolation)diff --git a/tests/kaggle/test_launch_cleanup.py b/tests/kaggle/test_launch_cleanup.pyindex e51f35c14d8..4f807fd10ae 100644--- a/tests/kaggle/test_launch_cleanup.py+++ b/tests/kaggle/test_launch_cleanup.py@@ -46,6 +46,14 @@ sys.path.insert(0, str(CI_DIR))import launch # noqa: E402+# One worker for this file. The tests below start a real launcher and signal it, then give it+# _DEATH_BUDGET_SEC to die; under `-n 4` with xdist's default scheduling four of them land on four+# workers of a four-core runner, and the budget stops measuring "did the handler run" and starts+# measuring "was the child ever scheduled". That is how CI produced a 120s timeout whose own+# faulthandler stack showed the child still parked on the stall line, having never reached its+# handler. Needs `--dist loadgroup`, which the jobs running this file pass.+pytestmark = pytest.mark.xdist_group(name = "kaggle_launch_signals")+class _StubKaggleApi:"""A client that can say WHICH account it is, because the real one can.