* Studio: defer llama.cpp update probes and self-heal MLX on macOS Two macOS startup problems shared one root area in the FastAPI lifespan: - The llama.cpp capability + freshness probes ran inline before the server yielded, so a cold/slow/flaky network on the GitHub freshness check blocked 'Application startup complete' (~34s on CI, longer in the field). Move both probes to a daemon thread; app.state stays None until ready (status routes already re-probe at request time). Opt out with UNSLOTH_DISABLE_UPDATE_CHECK=1. - Train and Export were greyed out because mlx/mlx-lm/mlx-vlm arrive only transitively and a resolver backtrack silently drops them, so CHAT_ONLY stayed true. Add utils/mlx_repair.py: when Apple Silicon is detected without MLX, reinstall mlx/mlx-lm/mlx-vlm by name on a daemon thread and re-run hardware detection (opt out UNSLOTH_DISABLE_MLX_AUTOREPAIR=1). Surface a chat_only_reason in /api/health plus a sidebar tooltip so a greyed Train/Export explains itself instead of failing silently. * Studio: guard model defaults against a None model name load_model_defaults(None) called model_name.lower() with no guard, raising 'Error loading model defaults for None' before any model is selected. Return an empty dict for a falsy/non-str name. * Studio: drop obsolete upstream macOS + Windows Blackwell prebuilt pins Both pins worked around gaps in ggml-org upstream prebuilts, but Studio now routes every GPU host and all of macOS to the unslothai/llama.cpp fork (published_repo_for_host), which ships the needed bundles, so both pins are dead code on the default install path: - macOS b9415: macOS always routes to the fork (its own macOS bundles), and host_supports_macos_minos() is the backstop. The pin only fired under an explicit --published-repo ggml-org override. - Windows Blackwell b9360: Windows-NVIDIA routes to the fork, whose windows-x64-cuda13 bundle covers Blackwell (manifest max_sm 120, toolkit 13.3), so the pin's self-disable check makes it dormant on every default install; it could only activate under the same upstream override on a 13.0-13.2 driver. Remove the pin constants, functions, and call sites. Keep the Blackwell capability detection (_drop_blackwell_incapable_windows_cuda, _host_is_blackwell, _windows_cuda_attempt_covers_blackwell) that still drops a non-sm_120 cuda-12.4 build on a Blackwell host. After this, an explicit --published-repo ggml-org override on a Blackwell 13.0-13.2 host loses its GPU fallback and lands on CPU; the default fork path is unaffected. Update the install selection-logic and macOS-compat unit tests for the new no-pin behavior. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Studio: walk back deeper on the macOS upstream prebuilt path After removing the b9415 macOS pin, the explicit --published-repo ggml-org upstream path still used the default 2-release fallback, so a pre-macOS-26 host behind a run of macOS-26-only builds would exhaust two too-new plans (minos is only checked post-download) and drop to a source build before reaching a loadable older release. Walk back as deep as the fork macOS path (DEFAULT_MAX_MACOS_RELEASE_FALLBACKS), turning the removed static pin into dynamic discovery. Addresses review feedback on the macOS upstream fallback. * Studio: pin transformers during MLX self-heal so it cannot break Studio mlx-lm/mlx-vlm declare transformers>=5, but the single-env install pins transformers==4.57.6. The self-heal used --upgrade with no constraint, so it could upgrade transformers in the live venv and break the rest of Studio just to make import mlx.core pass. Pin transformers to the installed version via a constraint file: the resolver either finds an mlx build compatible with it or fails (we stay chat-only), never upgrading transformers underneath Studio. Addresses review feedback on the MLX repair install. * Studio: harden MLX self-heal against an unsupported mlx-vlm Pinning transformers alone made uv backtrack mlx-vlm to 0.3.9 (below unsloth-zoo's mlx-vlm>=0.4.4), which imports but breaks VLM Train/Export -- so the self-heal could clear chat-only onto a broken stack. Mirror the main installer: set UV_OVERRIDE=overrides-darwin-arm64.txt so a current mlx-vlm coexists with the transformers pin, require the same minimum versions unsloth-zoo declares, and gate/validate on a full mlx_stack_available() check (not a bare import) so an old or partial stack stays chat-only. Addresses PR review. * Studio: filter Blackwell-incapable CUDA in resolve_upstream_asset_choice resolve_upstream_asset_choice returned the first windows-cuda choice unfiltered, so a Blackwell host could be handed an sm_120-incapable cuda-12.4 build while the sibling planners drop it. Apply _drop_blackwell_incapable_windows_cuda here too and fall through to the CPU bundle on a Blackwell host with no capable GPU asset. Addresses PR review. * Studio: re-poll health so MLX self-heal reaches an open UI The sidebar cached the initial /api/health, so a successful background MLX self-heal (chat_only flips false) did not re-enable Train/Export until a manual reload. While chat-only for the recoverable mlx_unavailable reason, re-poll /api/health and stop once Train/Export become available. Addresses PR review. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Studio: make the disabled Train/Export tooltip reachable The greyed Train/Export items pass a tooltip explaining why (e.g. MLX missing), but a disabled <button> fires no pointer events and SidebarMenuButton only showed tooltips while collapsed, so the explanation never appeared. Wrap a disabled button in a focusable span and show its tooltip while expanded too; enabled items keep the collapsed-only behavior. Addresses PR review. * Studio: gate Train/Export on the full MLX stack, not bare mlx.core detect_hardware enabled MLX training whenever `import mlx.core` worked, but the MLX self-heal (utils/mlx_repair) treats a stack without mlx-lm/mlx-vlm at the versions unsloth-zoo requires as inadequate. That asymmetry let the UI enable Train/Export on exactly the partial/backtracked stack the self-heal is trying to repair (greyed-in-but-broken VLM export). Gate on the same mlx_stack_available() criterion so a partial stack stays chat-only (reason mlx_unavailable) and the background repair restores it. Addresses PR review. * Fix MLX repair and health auth for PR #6494 * Fix macOS upstream prebuilt fallback for PR #6494 * Fix MLX stack validation for PR #6494 * Fix MLX self-heal validation for PR #6494 * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Review fixes: isolate hardware-state test, robust transformers pin - test_chat_only_reason.py: detect_hardware() assigns module globals directly, which monkeypatch does not revert; the autouse fixture now saves and restores DEVICE/CHAT_ONLY/CHAT_ONLY_REASON/IS_ROCM so a chat-only verdict here cannot leak into other backend tests (e.g. test_utils.py) on a GPU host. - mlx_repair.py: read the transformers version from importlib.metadata instead of importing transformers, so the install pin is not silently dropped when transformers has valid metadata but fails to import. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix CI: model full MLX stack in dispatch tests, keep selection test offline dispatch (macOS) job: - detect_hardware now gates MLX on the full stack (mlx_stack_available imports mlx_lm/mlx_vlm and checks dist versions), so faking only mlx.core makes the apple_silicon_mlx profile resolve to CPU. The dispatch tests assert the routing decision when the stack IS usable, so model a complete stack: test_hardware_dispatch_matrix patches utils.mlx_repair.mlx_stack_available and test_is_mlx_dispatch_gate patches hardware._has_usable_mlx_stack. The stack predicate's own internals stay covered by test_mlx_repair.py. Repo tests (CPU) job: - test_no_cuda_attempt_on_published_path_for_13_1 fell through to a live github_release_assets() upstream fetch after the Blackwell filter dropped every published attempt, which the offline security scanner blocks. Stub that fetch so the walk-back deterministically finds no usable CUDA build and raises PrebuiltFallback without network. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Harden MLX self-heal: prepare transformers constraint inside the try attempt_mlx_repair runs on a daemon thread, but _transformers_constraint_args was called before the try. A failure there (e.g. tempfile.mkstemp on a full disk or a bad TMPDIR) would propagate unhandled and silently kill the self-heal thread. Move the call inside the try and initialize constraint_path so any such failure is caught and leaves Studio chat-only instead of crashing the thread. --------- Co-authored-by: Daniel Han <michaelhan2050@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: wasimysaid <wasimysdev@gmail.com>
278 lines
9.5 KiB
Python
278 lines
9.5 KiB
Python
# SPDX-License-Identifier: AGPL-3.0-only
|
|
# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
|
|
|
|
"""MLX self-heal: on Apple Silicon with MLX missing, reinstall it by name on a
|
|
background thread (off the startup critical path). No-op elsewhere / when present
|
|
/ when disabled. Models on core.training.worker's runtime backend self-heal.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import sys
|
|
from pathlib import Path
|
|
|
|
import pytest
|
|
|
|
_BACKEND = Path(__file__).resolve().parent.parent
|
|
if str(_BACKEND) not in sys.path:
|
|
sys.path.insert(0, str(_BACKEND))
|
|
|
|
import utils.mlx_repair as mr # noqa: E402
|
|
|
|
|
|
@pytest.fixture(autouse = True)
|
|
def _reset_attempt_guard(monkeypatch):
|
|
monkeypatch.setattr(mr, "_attempted", False)
|
|
monkeypatch.delenv(mr.DISABLE_ENV_VAR, raising = False)
|
|
|
|
|
|
def test_uv_cmd_targets_this_interpreter_with_mlx_packages(monkeypatch):
|
|
monkeypatch.setattr(mr, "_uv_executable", lambda: "/usr/bin/uv")
|
|
cmd = mr._uv_install_cmd("--upgrade", *mr.MLX_PACKAGES)
|
|
assert cmd is not None
|
|
assert cmd[:5] == ["/usr/bin/uv", "pip", "install", "--python", sys.executable]
|
|
assert set(mr.MLX_PACKAGES) <= set(cmd)
|
|
# Minimum versions are pinned so the resolver cannot backtrack to an old
|
|
# mlx-vlm that imports but breaks VLM Train/Export.
|
|
assert "mlx-vlm>=0.4.4" in cmd
|
|
|
|
|
|
def test_uv_executable_finds_installer_location_when_path_is_minimal(monkeypatch, tmp_path):
|
|
uv = tmp_path / ".local" / "bin" / "uv"
|
|
uv.parent.mkdir(parents = True)
|
|
uv.write_text("#!/bin/sh\n", encoding = "utf-8")
|
|
uv.chmod(0o755)
|
|
monkeypatch.setattr(mr.shutil, "which", lambda _x: None)
|
|
monkeypatch.setattr(mr.Path, "home", lambda: tmp_path)
|
|
assert mr._uv_executable() == str(uv)
|
|
|
|
|
|
def test_no_uv_repair_stays_chat_only_without_pip(monkeypatch):
|
|
monkeypatch.setattr(mr, "_uv_executable", lambda: None)
|
|
monkeypatch.setattr(mr, "_transformers_constraint_args", lambda: ([], None))
|
|
called = {"run": False}
|
|
|
|
def _fake_run(*_args, **_kwargs):
|
|
called["run"] = True
|
|
raise AssertionError("plain pip fallback must not run")
|
|
|
|
monkeypatch.setattr(mr.subprocess, "run", _fake_run)
|
|
assert mr.attempt_mlx_repair() is False
|
|
assert called["run"] is False
|
|
|
|
|
|
def test_constraint_pins_installed_transformers(monkeypatch):
|
|
transformers = pytest.importorskip("transformers")
|
|
args, path = mr._transformers_constraint_args()
|
|
try:
|
|
assert args[:1] == ["--constraint"]
|
|
assert args[1] == path
|
|
assert Path(path).read_text().strip() == f"transformers=={transformers.__version__}"
|
|
finally:
|
|
if path:
|
|
Path(path).unlink(missing_ok = True)
|
|
|
|
|
|
def test_repair_install_pins_transformers_and_cleans_up(monkeypatch):
|
|
pytest.importorskip("transformers")
|
|
captured = {}
|
|
created_paths = []
|
|
real_args = mr._transformers_constraint_args
|
|
|
|
def _spy_args():
|
|
args, path = real_args()
|
|
if path:
|
|
created_paths.append(path)
|
|
return args, path
|
|
|
|
monkeypatch.setattr(mr, "_transformers_constraint_args", _spy_args)
|
|
monkeypatch.setattr(mr, "_uv_executable", lambda: "/usr/bin/uv")
|
|
|
|
class _Result:
|
|
returncode = 0
|
|
stdout = ""
|
|
|
|
def _fake_run(cmd, **kwargs):
|
|
captured["cmd"] = cmd
|
|
captured["env"] = kwargs.get("env")
|
|
return _Result()
|
|
|
|
monkeypatch.setattr(mr.subprocess, "run", _fake_run)
|
|
monkeypatch.setattr(mr, "mlx_stack_available", lambda: True)
|
|
|
|
assert mr.attempt_mlx_repair() is True
|
|
cmd = captured["cmd"]
|
|
# transformers is pinned via a constraint file so the mlx install cannot
|
|
# upgrade it underneath Studio, and the temp constraint file is cleaned up.
|
|
assert "--constraint" in cmd
|
|
assert "--upgrade" in cmd
|
|
reinstall_pairs = set(zip(cmd, cmd[1:]))
|
|
for name in mr._MLX_PACKAGE_NAMES:
|
|
assert ("--reinstall-package", name) in reinstall_pairs
|
|
for pkg in mr.MLX_PACKAGES:
|
|
assert pkg in cmd
|
|
assert created_paths and not Path(created_paths[0]).exists()
|
|
# The install mirrors the main installer by relaxing the transformers pin via
|
|
# UV_OVERRIDE so a current mlx-vlm can coexist with transformers==4.57.6.
|
|
env = captured["env"]
|
|
assert env is not None
|
|
assert env.get("UV_OVERRIDE", "").endswith("overrides-darwin-arm64.txt")
|
|
|
|
|
|
def test_repair_rejects_inadequate_stack(monkeypatch):
|
|
# A successful uv run that still leaves an old/missing mlx-vlm must NOT clear
|
|
# chat-only: attempt_mlx_repair returns False so Train/Export stay disabled.
|
|
class _Result:
|
|
returncode = 0
|
|
stdout = ""
|
|
|
|
monkeypatch.setattr(mr.subprocess, "run", lambda *a, **k: _Result())
|
|
monkeypatch.setattr(mr, "mlx_stack_available", lambda: False)
|
|
assert mr.attempt_mlx_repair() is False
|
|
|
|
|
|
def test_repair_invalidates_import_caches_before_stack_check(monkeypatch):
|
|
events = []
|
|
|
|
class _Result:
|
|
returncode = 0
|
|
stdout = ""
|
|
|
|
def _stack_available():
|
|
events.append("check")
|
|
assert events == ["invalidate", "check"]
|
|
return True
|
|
|
|
monkeypatch.setattr(mr.subprocess, "run", lambda *a, **k: _Result())
|
|
monkeypatch.setattr(mr, "_uv_executable", lambda: "/usr/bin/uv")
|
|
monkeypatch.setattr(mr, "_transformers_constraint_args", lambda: ([], None))
|
|
monkeypatch.setattr(mr.importlib, "invalidate_caches", lambda: events.append("invalidate"))
|
|
monkeypatch.setattr(mr, "mlx_stack_available", _stack_available)
|
|
|
|
assert mr.attempt_mlx_repair() is True
|
|
assert events == ["invalidate", "check"]
|
|
|
|
|
|
def test_stack_unavailable_without_mlx(monkeypatch):
|
|
import importlib.metadata as metadata
|
|
|
|
def _missing(_name):
|
|
raise metadata.PackageNotFoundError(_name)
|
|
|
|
monkeypatch.setattr(metadata, "version", _missing)
|
|
assert mr.mlx_stack_available() is False
|
|
|
|
|
|
def test_stack_unavailable_checks_versions_before_imports(monkeypatch):
|
|
import importlib.metadata as metadata
|
|
|
|
def _version(name):
|
|
if name == "mlx":
|
|
return "0.21.0"
|
|
return mr._MLX_MIN_VERSIONS[name]
|
|
|
|
def _import_module(_name):
|
|
raise AssertionError("MLX modules must not import before versions pass")
|
|
|
|
monkeypatch.setattr(metadata, "version", _version)
|
|
monkeypatch.setattr(mr.importlib, "import_module", _import_module)
|
|
assert mr.mlx_stack_available() is False
|
|
|
|
|
|
def test_stack_unavailable_when_companion_import_fails(monkeypatch):
|
|
import importlib.metadata as metadata
|
|
|
|
monkeypatch.setattr(metadata, "version", lambda name: mr._MLX_MIN_VERSIONS[name])
|
|
|
|
def _import_module(name):
|
|
if name == "mlx_vlm":
|
|
raise ModuleNotFoundError(name)
|
|
return object()
|
|
|
|
monkeypatch.setattr(mr.importlib, "import_module", _import_module)
|
|
assert mr.mlx_stack_available() is False
|
|
|
|
|
|
def test_stack_available_requires_runtime_imports_and_versions(monkeypatch):
|
|
import importlib.metadata as metadata
|
|
|
|
imported = []
|
|
|
|
def _import_module(name):
|
|
imported.append(name)
|
|
return object()
|
|
|
|
monkeypatch.setattr(mr.importlib, "import_module", _import_module)
|
|
monkeypatch.setattr(metadata, "version", lambda name: mr._MLX_MIN_VERSIONS[name])
|
|
|
|
assert mr.mlx_stack_available() is True
|
|
assert imported == list(mr._MLX_RUNTIME_IMPORTS)
|
|
|
|
|
|
def test_no_op_off_apple_silicon(monkeypatch):
|
|
monkeypatch.setattr(mr, "is_apple_silicon", lambda: False)
|
|
called = {"n": 0}
|
|
monkeypatch.setattr(
|
|
mr, "attempt_mlx_repair", lambda **_k: called.__setitem__("n", called["n"] + 1) or True
|
|
)
|
|
assert mr.start_mlx_autorepair_if_needed() is False
|
|
assert called["n"] == 0
|
|
|
|
|
|
def test_no_op_when_mlx_stack_present(monkeypatch):
|
|
monkeypatch.setattr(mr, "is_apple_silicon", lambda: True)
|
|
monkeypatch.setattr(mr, "mlx_stack_available", lambda: True)
|
|
started = mr.start_mlx_autorepair_if_needed()
|
|
assert started is False
|
|
|
|
|
|
def test_disable_env_skips(monkeypatch):
|
|
monkeypatch.setattr(mr, "is_apple_silicon", lambda: True)
|
|
monkeypatch.setattr(mr, "mlx_stack_available", lambda: False)
|
|
monkeypatch.setenv(mr.DISABLE_ENV_VAR, "1")
|
|
assert mr.start_mlx_autorepair_if_needed() is False
|
|
|
|
|
|
def test_apple_silicon_missing_mlx_starts_repair_and_redetects(monkeypatch):
|
|
import threading
|
|
|
|
monkeypatch.setattr(mr, "is_apple_silicon", lambda: True)
|
|
monkeypatch.setattr(mr, "mlx_stack_available", lambda: False)
|
|
|
|
repaired = {"called": False}
|
|
|
|
def _fake_repair(**_kw):
|
|
repaired["called"] = True
|
|
return True
|
|
|
|
redetected = {"called": False}
|
|
|
|
# _run_repair_and_redetect imports utils.hardware.hardware lazily; stub repair
|
|
# and capture that re-detection is invoked on success.
|
|
monkeypatch.setattr(mr, "attempt_mlx_repair", _fake_repair)
|
|
|
|
import utils.hardware.hardware as hw
|
|
|
|
monkeypatch.setattr(hw, "detect_hardware", lambda: redetected.__setitem__("called", True))
|
|
|
|
started = mr.start_mlx_autorepair_if_needed()
|
|
assert started is True
|
|
|
|
# Join the daemon thread deterministically.
|
|
for thread in threading.enumerate():
|
|
if thread.name == "mlx-autorepair":
|
|
thread.join(timeout = 5)
|
|
|
|
assert repaired["called"] is True
|
|
assert redetected["called"] is True
|
|
|
|
|
|
def test_attempts_only_once_per_process(monkeypatch):
|
|
monkeypatch.setattr(mr, "is_apple_silicon", lambda: True)
|
|
monkeypatch.setattr(mr, "mlx_stack_available", lambda: False)
|
|
monkeypatch.setattr(mr, "attempt_mlx_repair", lambda **_k: False)
|
|
|
|
first = mr.start_mlx_autorepair_if_needed()
|
|
second = mr.start_mlx_autorepair_if_needed()
|
|
assert first is True
|
|
assert second is False # guard prevents a second concurrent attempt
|