unsloth/tests/studio/install/test_rocm_support.py
Leo Borcherding b6d5636cc0
fix/strix halo and windows AMD ROCm support (#5301)
* fix(studio): set HIP_VISIBLE_DEVICES in apply_gpu_ids for ROCm training workers

Training workers are spawned via multiprocessing spawn before detect_hardware()
runs, so IS_ROCM is still False. If the user never set HIP_VISIBLE_DEVICES in
their shell, _inherits_rocm_visibility is also False, leaving the worker with
only CUDA_VISIBLE_DEVICES set. On ROCm hosts the HIP runtime honors
HIP_VISIBLE_DEVICES over CUDA_VISIBLE_DEVICES, so the worker saw the full
device list and torch raised "no usable HIP accelerator" on some setups.

Fall back to probing torch.version.hip (a build-time attribute, safe to read
before GPU init) to detect ROCm when neither IS_ROCM nor inherited env vars
are available. Mirrors the existing fix in llama_cpp.py for llama-server
subprocess GPU pinning.

Fixes https://github.com/unslothai/unsloth/issues/5180

* test: tighten apply_gpu_ids ROCm fallback assertions

Replace loose OR chain with exact string matches, split into three
focused tests, and add a guard check for the try/except wrapper.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: detect ROCm unified memory (Strix Halo / AMD iGPU) via torch fallback

amd-smi on iGPUs with shared/unified memory (e.g. Radeon 8060S on Strix
Halo) reports only the dedicated VRAM slice (~512 MB) in its metric output,
so get_visible_gpu_utilization() was returning usable_gb ≈ 0.35 GB instead
of the full GTT pool (~128 GB).  torch.cuda.mem_get_info() already surfaces
the correct unified-pool size.

Add _reconcile_rocm_unified_memory(): after amd-smi returns a valid result
on a ROCm device, cross-check each device's vram_total_gb against
torch.cuda.mem_get_info().  When torch reports a larger total, replace the
amd-smi VRAM fields in-place.  No-op for discrete AMD GPUs where the two
sources agree.

Fixes: "Falling back to all visible GPUs -- model may not fit" on AMD iGPU
machines even when 100+ GB of unified memory is available.

* Apply unified-memory reconciliation in get_gpu_utilization too

The visible-GPU path was already corrected for AMD iGPUs with unified memory
(Strix Halo / Radeon 8060S), but get_gpu_utilization was still returning the
raw 512 MB amd-smi VRAM slice. Studio's /api/train/hardware endpoint and the
live GPU monitor read from this primary path, so users continued seeing the
wrong total even after auto_select_gpu_ids picked the right device.

Refactor to share the per-device correction:
  * _apply_unified_memory_correction(metrics, torch_info) -- the actual
    replacement logic, in-place on a single metrics dict.
  * _reconcile_rocm_unified_memory(...)                   -- multi-device,
    iterates utilization["devices"] (visible-GPU path).
  * _reconcile_primary_rocm_unified_memory(...)           -- single flat
    metrics dict (primary-GPU path), uses parent_visible_spec to pick the
    primary index, falls back to ordinal 0 when no visibility env is set.

get_gpu_utilization now calls the primary reconciler under IS_ROCM, so both
endpoints surface the real unified-memory pool on iGPUs while leaving
discrete AMD GPUs untouched (torch_total <= smi_total -> no replace).

* Use 'is not None' and log debug on torch.version.hip probe failures

Two small follow-ups to the apply_gpu_ids ROCm fallback:

1. Match detect_hardware()'s 'getattr(torch.version, "hip", None) is not None'
   form so the entire codebase has one canonical 'this torch was built with
   HIP' check. On every shipping torch wheel hip is either None or a non-empty
   version string, so the new form agrees with the old bool() form on every
   real install.

2. Log the probe failure at debug level instead of swallowing it silently.
   The broad 'except Exception' is intentional (we never want apply_gpu_ids
   to crash a worker over a probe), but the silent pass made it impossible
   to tell whether the fallback was firing or being skipped.

* fix(studio): honour HIP_VISIBLE_DEVICES in _get_parent_visible_gpu_spec before IS_ROCM is set

When a user has HIP_VISIBLE_DEVICES set in their shell (e.g. "1" to select
GPU 1) but detect_hardware() has not yet run in the Studio parent process,
IS_ROCM is still False.  _get_parent_visible_gpu_spec() was gated on IS_ROCM
so it fell through to CUDA_VISIBLE_DEVICES (unset), saw all physical GPUs,
and auto-selected index 0.  apply_gpu_ids then overwrote HIP_VISIBLE_DEVICES
with "0", making the intended GPU invisible to ROCm torch in the worker,
which triggered the "no usable HIP accelerator" error (issue #5180).

Apply the same _inherits_rocm_visibility pattern already used in
apply_gpu_ids: check for HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES in the
environment regardless of IS_ROCM so the correct GPU index is preserved.

* fix(install): harden AMD ROCm GPU detection for multi-GPU and env-filtered setups

The previous rocminfo awk pattern could miss discrete GPUs on machines
where HIP_VISIBLE_DEVICES/ROCR_VISIBLE_DEVICES is used to mask an
integrated GPU — the env vars filter rocminfo output but may not
propagate into the install script subprocess, causing detection to
fail entirely.

Two changes:
- Tighten rocminfo pattern from /gfx[0-9]/ && !/gfx000/ to
  /gfx[1-9][0-9]/ — simpler and correctly excludes the CPU agent
  (gfx000) without a negative lookahead
- Add sysfs KFD topology fallback: reads
  /sys/class/kfd/kfd/topology/nodes/*/gpu_id which is a kernel-level
  view unaffected by HIP_VISIBLE_DEVICES or ROCR_VISIBLE_DEVICES

Fixes detection failure reported in Discord by Chains (gfx1201 + iGPU
machine where env var exclusion of the iGPU caused rocminfo to return
no usable device).

* Fix KFD sysfs awk fallback to read properties file

The fallback added by this PR reads /sys/class/kfd/kfd/topology/nodes/*/gpu_id
files but matches the literal token 'gpu_id' against their content. Those
files contain only a single decimal value (e.g. '0' for CPU agents, '50432'
for GPU agents), so the regex never matches and 'found' stays 0, making the
fallback a no-op on every host. The properties file in the same directory
contains key/value lines like 'gpu_id 50432' which is what the existing awk
pattern expects.

Reproduced with a synthetic sysfs layout: against gpu_id files awk exits 1;
against properties files awk exits 0 when any node reports gpu_id > 0.

* fix(setup.ps1): detect AMD ROCm GPU on Windows, bring to parity with setup.sh

setup.ps1 only checked nvidia-smi and fell straight to "gpu: none" on AMD
machines. setup.sh already probed rocminfo/amd-smi/hipconfig/hipinfo.

Add three-tier detection mirroring install_llama_prebuilt.py's detect_host():
1. hipinfo: gcnArchName in output confirms a real HIP GPU (not just SDK)
2. amd-smi list: "GPU: <digit>" data rows as fallback
3. WMI Win32_VideoController: last resort -- detects AMD GPU even without
   HIP SDK, then guides user to install it rather than silently going CPU

Also corrects the "none" message to mention AMD ROCm alongside NVIDIA so
users with AMD hardware understand the requirement.

Fixes: rohit-style install where Strix Halo (Radeon 8060S) showed
"gpu: none" even with the HIP SDK present.

* fix(install.ps1): detect AMD ROCm GPU on Windows, bring to parity with setup.ps1

install.ps1 had the same nvidia-smi-only GPU detection as setup.ps1 before
the setup.ps1 fix. Applies the same three-tier AMD detection:
1. hipinfo: gcnArchName confirms real HIP GPU
2. amd-smi list: GPU data rows as fallback
3. WMI Win32_VideoController: detects AMD GPU without HIP SDK and guides
   user to install it

Fixes: install.ps1 showing "gpu: none" while setup.ps1 correctly showed
"AMD GPU detected" on the same machine (reported by rohit, RX 7600 XT).

* fix(install.ps1): suppress 'No NVIDIA GPU detected' when AMD GPU is present

* feat: add Windows AMD ROCm PyTorch wheel installation

install_python_stack.py:
- Add _ROCM_WINDOWS_WHEEL_BASE and _ROCM_WINDOWS_RELEASES constants
  pointing to AMD repo.radeon.com (ROCm 7.2 -> torch 2.9.1+rocm7.2.1)
- Extend _ensure_rocm_torch() with a Windows branch: detects ROCm via
  _has_rocm_gpu() / _detect_rocm_version(), requires Python 3.12 (cp312
  is the only ABI AMD publishes for Windows), installs the direct wheel
  URL from repo.radeon.com

install.ps1:
- Capture ROCmVersion during AMD detection via hipconfig --version /
  amd-smi version (needed for wheel URL selection)
- After Get-TorchIndexUrl, add an AMD wheel override block: when HasROCm
  and Python 3.12 detected, set ROCmTorchWheelUrl to AMD wheel URL
- Expand torch install branch to handle ROCmTorchWheelUrl with
  uv pip install --force-reinstall --no-cache-dir

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: also install torchvision and torchaudio from AMD Windows repo

AMD publishes matching torchvision-0.24.1+rocm7.2.1 and
torchaudio-2.9.1+rocm7.2.1 cp312 wheels at the same repo.radeon.com
release folder. Install all three in both install.ps1 and
install_python_stack.py Windows ROCm path.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* feat: add ROCm 7.1.1 Windows wheel mapping

AMD uses a different version string for 7.1.1 wheels:
2.9.0+rocmsdk20251116 (date-tagged) instead of +rocm7.1.1.
Adds the 7.1.1 release folder to both install.ps1 and
install_python_stack.py so users with ROCm 7.1 get ROCm
torch instead of falling back to CPU.

* fix: install rocm_sdk_core and rocm_sdk_libraries_custom alongside torch

The AMD Windows torch wheels declare rocm[libraries]==<ver> as a hard
dependency. Without installing rocm_sdk_core and rocm_sdk_libraries_custom
from the same AMD release folder, uv cannot resolve the dependency and
fails with 'No solution found'. Include all 5 wheels in one install call.

* fix: expand ROCm wheel array to scalars for Invoke-InstallCommand

@array splatting inside a scriptblock only works when the native command
is prefixed with '&'. Invoke-InstallCommand uses '& $Command' to run the
block, so @ROCmAllWheelUrls was not being expanded. Extract to scalar
variables $rw0-$rw4 which are captured correctly by the closure.

* fix: use --no-deps for AMD Windows torch wheel install

uv's resolver looks up rocm[libraries]==0.1.dev0 on PyPI during
dependency resolution before downloading any wheels, and fails because
the package doesn't exist on PyPI. --no-deps skips resolution entirely
and installs all 5 AMD wheels directly. The GPU runtime dependency is
satisfied by the HIP SDK, not a Python package.

* fix: setup.ps1 and install_python_stack.py now install ROCm torch on Windows

setup.ps1 was always setting CuTag='cpu' for non-NVIDIA hosts and installing
cpu-only PyTorch, overwriting the ROCm torch installed by install.ps1.
Adds the same AMD wheel selection logic (ROCm version detection, Python 3.12
check, 5-wheel install with --no-deps) to setup.ps1's torch install block.

install_python_stack.py: remove IS_WINDOWS guard from _ensure_rocm_torch()
call site so the Windows path in _ensure_rocm_torch() is reachable during
'unsloth studio update' as well.

* fix: suppress manual-install warning when ROCm torch already present; fix progress counter

- Gate the 'must be installed manually' warning on torch.version.hip being empty
  so it doesn't fire when our ROCm torch install succeeded
- Update _TOTAL counter to include the 3 ROCm steps on Windows now that
  _ensure_rocm_torch() is called there (fixes 10/9 display)

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* feat: add rocm step display in setup.ps1; fix warning and progress counter

- Add 'rocm' step after 'cuda' in setup.ps1 showing ROCm version or HIP SDK missing
- Move ROCm version detection up to GPU detection block so it's available early
- Suppress 'must be installed manually' warning when torch.version.hip is set
- Fix _TOTAL counter to include ROCm steps on Windows (fixes 10/9 display)

* fix: detect AMD SDK ROCm torch via __version__ when torch.version.hip is unset

AMD's repo.radeon.com wheels (e.g. 2.9.0+rocmsdk20251116) do not set
torch.version.hip, leaving it None. All three probes that relied solely on
torch.version.hip now also check for 'rocm' in torch.__version__.lower():

- hardware.py detect_hardware(): IS_ROCM was never set, causing the studio
  to report 'Hardware detected: CPU' even after AMD wheels were installed
  and HIP DLLs were on PATH.
- install_python_stack.py _ensure_rocm_torch(): skip-if-already-installed
  probe would always reinstall on subsequent runs.
- install_python_stack.py Windows AMD warning: suppression check always
  failed, so the 'must be installed manually' note kept appearing after
  a successful AMD wheel install.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* perf: drop --no-cache-dir from AMD ROCm torch wheel installs

uv caches downloaded wheels by default; passing --no-cache-dir forced a
full redownload of the ~2 GB torch wheel on every install run. CUDA installs
never had this flag -- AMD was the only path affected.

* fix: use install-state flag instead of subprocess probe for AMD Windows warning

Replace the subprocess torch probe in the post-install warning block with a
module-level _rocm_windows_torch_installed flag set by _ensure_rocm_torch().
Subprocess re-import of torch is unnecessary and fragile -- the install
function already knows whether it succeeded.

* fix: hoist global declaration to top of _ensure_rocm_torch

Python requires the global statement to appear before any assignment
to the variable within a function. Moving it to the function top fixes
the SyntaxError on line 354.

* fix: pass AMD torch install status via env var to suppress false warning

setup.ps1 now sets UNSLOTH_ROCM_TORCH_INSTALLED=1 after a successful AMD
wheel install. install_python_stack.py reads this at the top of
_ensure_rocm_torch() to skip both the subprocess probe and the warning --
no re-import of torch needed, and the warning message now correctly says
'could not be auto-installed' rather than 'must be installed manually'.

* fix: register ROCm DLL directory before torch import on Windows

Python 3.8+ ignores PATH for extension DLL loading on Windows; amdhip64.dll
and other HIP runtime DLLs must be registered via os.add_dll_directory().
Without this, torch.cuda.is_available() always returns False on AMD ROCm
Windows even when HIP_PATH is correctly set in system environment variables.

Reads HIP_PATH / ROCM_PATH env vars first, then falls back to scanning
common ROCm install roots (C:\Program Files\AMD\ROCm, F:\ROCm, C:\ROCm).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: remove hardcoded non-standard ROCm paths from DLL directory scan

Only use HIP_PATH/ROCM_PATH (set by AMD installer) and the standard
C:\Program Files\AMD\ROCm\<version>\bin location. Custom drive paths
like F:\ROCm are user-specific and should not be hardcoded.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: prevent torchao overrides step from overwriting AMD ROCm torch

torchao==0.14.0 in overrides.txt declares torch as a dependency. Without
--no-deps, uv resolves torch from PyPI and installs 2.11.0+cpu on top of
the AMD ROCm wheels (2.9.0+rocmsdk20251116). This was the root cause of
'Hardware detected: CPU' -- the AMD wheels were installed but then
immediately overwritten by the overrides step.

When _rocm_windows_torch_installed is True, add --no-deps to the overrides
pip_install call so torchao is installed without pulling in CPU torch.

* fix: add rocm_sdk namespace tarball to Windows ROCm wheel installs

torch/_rocm_init.py calls `import rocm_sdk` at startup, which requires
the rocm namespace tarball (rocm-*.tar.gz) in addition to the SDK wheel
packages. This tarball was missing from both install.ps1 and setup.ps1,
causing ModuleNotFoundError on first torch import.

- Add rocm-0.1.dev0.tar.gz to ROCm 7.1.1 install (provides rocm_sdk namespace)
- Add rocm-7.2.1.tar.gz + rocm_sdk_devel to ROCm 7.2.1 install
- Install tarball in a dedicated step before main SDK/torch wheels
- Switch to @array splatting in install.ps1 scriptblock for dynamic wheel count
- Remove --no-cache-dir from Python-side ROCm wheel install (prevents ~2GB redownload)

* feat: enable ROCm 7.2 torch install + warn on gfx1151 with ROCm < 7.2

Chigoma333 (AMD Radeon 8060S / gfx1151, Strix Halo) confirmed that ROCm
7.1 segfaults when tensors are moved to GPU, but ROCm 7.2 + torch
2.11.0+rocm7.2 works fully including training.

Changes:
- Uncomment (7,2): "rocm7.2" in _ROCM_TORCH_INDEX (was blocked by <2.11.0)
- Add _ROCM_TORCH_PKG_SPECS dict with per-tag version bounds:
  rocm7.2 → torch>=2.11.0,<2.12.0; all older tags → <2.11.0
- Add _detect_amd_gfx_codes() helper that parses rocminfo output
- Warn on gfx1151/gfx1150 (Strix Halo) when ROCm < 7.2 is installed,
  pointing users at the known segfault and recommending upgrade
- install.sh get_torch_index_url(): enable rocm7.2 case (previously capped
  to rocm7.1), cap unknown future tags to rocm7.2
- install.sh: override TORCH_CONSTRAINT to >=2.11.0,<2.12.0 when rocm7.2
  index is selected, so pip can actually resolve torch 2.11.0

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: prefer Python 3.12 for AMD ROCm users when 3.13 is also installed

After GPU detection, if ROCm HIP SDK is found and the selected Python
is not 3.12, run a second pass to locate a 3.12 install via py.exe and
PATH (catches uv-managed installs). Switch $DetectedPython to 3.12 so
the venv is created with a compatible interpreter for the cp312-only AMD
Windows torch wheels.

NVIDIA and Intel GPU paths are unaffected -- the re-detection block only
runs when $HasROCm is true.

Fixes: #5301

* fix: also check uv-managed Python 3.12 for AMD ROCm #5301

* fix: hide amd-smi console popups on Windows, guard torch.distributed.is_initialized for ROCm #5301

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: suppress remaining console popups on Windows, patch torch.distributed.is_initialized for ROCm #5301

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: stub all missing torch.distributed attrs for ROCm Windows wheel #5301

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: inject torch.distributed stub when C backend missing in ROCm Windows wheel #5301

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(rocm/windows): pre-stub torch._C._distributed_c10d + raise amd-smi timeout

Two fixes for Windows ROCm regressions reported by electroglyph on #5301:

1. worker.py — torch.distributed stub now fires unconditionally on Windows
   The previous stub only injected sys.modules in the except branch, meaning
   it was silently skipped when `import torch.distributed` happened to succeed
   (the C backend is lazily resolved).  The crash then hit later when
   transformers/trl triggered the lazy load.  Fix: on win32 we pre-populate
   sys.modules['torch._C._distributed_c10d'] AND set the attribute on the
   torch._C extension module *before* attempting the import, covering both
   the early-ImportError and lazy-load failure modes.

2. amd.py — increase amd-smi timeout from 5 s to 30 s on Windows (10 s Linux)
   amd-smi on Windows must cold-init the ROCm runtime on first invocation;
   5 s was consistently too short, producing repeated 'Command timed out'
   warnings in the server log.  30 s gives enough headroom without blocking
   indefinitely on broken installs.

3. install.ps1 — widen Python 3.12 enforcement to ROCmGpuLabel (WMI-only path)
   Users whose HIP SDK is not on PATH were detected via WMI but not switched
   to Python 3.12 before the install started, causing a second pass.  Guard
   now fires on (HasROCm -or ROCmGpuLabel).

* fix(rocm): guard c10d stub, fix TorchIndexFamily for 7.1, clean dead code + comments

- worker.py: wrap c10d stub injection in `if _c10d_key not in sys.modules` so
  Windows NVIDIA users with a real torch.distributed are never affected
- install.ps1: fix Get-TauriTorchIndexFamily receiving hardcoded "rocm7.2"
  even when ROCm 7.1 wheels are installed; now branches on $ROCmVersion
- main.py: remove dead `import ctypes as _ctypes` (ctypes is never called)
- hardware.py, install_python_stack.py, worker.py, install.ps1: shorten
  verbose multi-line comment blocks throughout
- tests: update 4 stale assertions that expected rocm7.2 to be absent/capped

* fix(tests): match windows AMD warning assertion to actual source string

* chore: trim verbose comment blocks across all ROCm-related files

* fix: guard reconcile call against None numeric_ids; add torchvision lower bounds

* fix(install.ps1): recreate venv with Python 3.12 after ROCm switch

Venv was created with 3.13 before GPU detection ran; switching
$DetectedPython to 3.12 had no effect since $VenvPython still
pointed to the 3.13 interpreter inside the already-created venv.

* ux: detect AMD GPU before Python selection to avoid double venv creation

- Early hipinfo + WMI probe runs before Find-CompatiblePython so Python
  3.12 is selected upfront when AMD is detected; venv is now created
  exactly once instead of 3.13 then immediately 3.12.
- Post-venv recreation block replaced with a simple warning for the rare
  case where AMD was missed by the early probe.
- setup.ps1: show venv's actual Python version (e.g. 3.12) instead of
  the system Python found by the pre-activation search (was showing 3.13).

* fix(rocm/win): auto-stub all _distributed_c10d symbols via PEP-562 __getattr__

The bare ModuleType stub caused ImportError when torch._dynamo was imported
(triggered by trainer.py accessing torch._dynamo.config at load time).
torch._dynamo pulls in torch.distributed.fsdp._flat_param which does:
  from torch._C._distributed_c10d import FakeProcessGroup
and potentially other symbols. Adding module __getattr__ auto-creates a
stub class for any missing symbol so all such imports succeed without
enumerating every individual symbol. Applied to both the primary stub
and the fallback stub in the except branch.

* chore: trim c10d stub comment

* fix(rocm/win): auto-stub missing torch.distributed attrs (Store, ProcessGroup, …)

* fix(rocm/win): pre-stub fsdp submodules in sys.modules; fix __getattr__ subpackage clash

* feat(rocm/win): arch-aware wheel selector always picks newest ROCm release

Replace HIP-SDK-version-gated wheel selection with GPU arch-based logic.
Select-ROCmWheelRelease (PS) and _select_windows_rocm_release (Python) map
gcnArchName → minimum ROCm version, then pick the newest available release
that satisfies it (currently always rocm-rel-7.2.1 for any supported GPU).
Wheels bundle their own ROCm runtime so the installed HIP SDK 7.1 does not
prevent using 7.2.1 wheels on gfx1200 (RX 9060 XT) and similar RDNA 4 GPUs.

Also installs the bitsandbytes Windows ROCm continuous-release wheel and sets
BNB_ROCM_VERSION=72 in worker.py before ML imports so bnb loads the
libbitsandbytes_rocm72.dll that ships in that wheel.

* fix(rocm/win): stub class metaclass for ProcessGroup.BackendType; amd-smi circuit breaker

torchao.float8.inference accesses ProcessGroup.BackendType as a class-level
attribute.  Plain type() stubs have no __getattr__ on the metaclass so this
raises AttributeError.  Introduce _StubClassMeta whose __getattr__ returns
child stub classes, fixing the torchao import chain.

Add an amd-smi circuit breaker in amd.py: after 3 consecutive failures the
module stops spawning the process, eliminating the repeated Windows UAC /
DiskPart elevation prompts caused by polling a non-functional amd-smi.

Also guard BNB_ROCM_VERSION=72 behind a DLL existence check so bitsandbytes
fails with its own detection message rather than a harder "DLL not found" when
the Windows ROCm bnb wheel is not yet installed.

* fix: stub __members__ so torchao float8 enum check doesn't crash on ROCm Windows

torchao.float8.inference accesses ProcessGroup.BackendType.__members__
expecting a Python Enum registry dict. _StubClassMeta.__getattr__ was
blocking all dunder attributes, causing AttributeError. Return {} for
__members__ specifically so the isinstance/iteration checks pass cleanly.

* fix: stub distributed tensor/functional_collectives to prevent missing C++ op crash on ROCm Windows

torch._dynamo.trace_rules eagerly loads torch.distributed.tensor at import
time, which pulls in _functional_collectives.py. That file registers Meta
kernels for _c10d_functional C++ ops, but those ops are only registered
by torch._C._distributed_c10d — a C extension absent from ROCm Windows
wheels. Pre-stubbing the affected modules in sys.modules prevents the real
import chain from running and avoids the "operator does not exist" crash.

* fix: give mod stubs __path__ and pre-stub _tensor to fix 'not a package' import error

_make_mod_stub now sets __path__=[] so Python treats stub modules as
packages. Without it, any import of a submodule raises "is not a package".
Also pre-stub torch.distributed._tensor and its submodules so that
_tensor/__init__.py (which re-exports from torch.distributed.tensor) never
runs and torchao's `from torch.distributed._tensor import DTensor` gets a
harmless stub instead of crashing.

* fix: stub torch.ops._c10d_functional namespace with hashable op sentinels

torchao.dtypes.nf4tensor uses _c10d_functional ops as dict keys at import
time (all_gather_into_tensor.default, wait_tensor.default) and
torch.ops.c10d.scatter_.default. None of these ops are registered on ROCm
Windows because torch._C._distributed_c10d (the C extension) doesn't ship.
Replace the whole _c10d_functional namespace with a custom stub whose ops
return hashable .default objects, so dict-key construction doesn't crash.
Also inject a scatter_ stub into torch.ops.c10d if it's missing.

* fix: stub entire torchao package on ROCm Windows instead of individual ops

torchao is not supported on ROCm Windows and its import chain transitively
requires torch._C._distributed_c10d (absent from the ROCm Windows wheel).
Rather than stub each missing op one by one, stub the whole torchao package
upfront. Unsloth uses bitsandbytes for quantization, not torchao, so this
has no functional impact. transformers gracefully handles an importable-but-
empty torchao by disabling TorchAoHfQuantizer.

* fix: set __spec__ on mod stubs so importlib.util.find_spec doesn't raise

Manually-injected sys.modules entries have __spec__=None by default.
importlib.util.find_spec() raises ValueError when it finds a module in
sys.modules with __spec__=None (transformers.utils.import_utils hits this
when checking if torchao is available). Give every stub a minimal
ModuleSpec(name, loader=None, is_package=True) to satisfy find_spec.

* fix: add meta path finder to auto-stub subpackages of stub modules

`import torchao.prototype` goes through the import machinery, not
__getattr__, so an empty __path__ means ModuleNotFoundError. Rather than
list every submodule explicitly, register a MetaPathFinder that intercepts
any import whose parent is one of our stubs (detected by loader=None in the
parent's ModuleSpec). Real installed packages always have a SourceFileLoader
so they are never intercepted. Also register child stubs in sys.modules
from __getattr__ as a belt-and-suspenders measure.

* fix: use _unsloth_stub sentinel instead of loader=None for stub detection

The import machinery overwrites module.__spec__ with the spec returned by
find_spec (which has loader=_StubSubpackageLoader, not None), so the
loader=None check broke for second-level subpackages. Switch to a custom
_unsloth_stub object identity sentinel set directly on each stub module --
it survives __spec__ being replaced and correctly identifies stubs at any
depth (torchao.prototype.safetensors, etc.).

* refactor(rocm/win): switch to repo.amd.com arch-aware index, remove stubs

AMD recommends repo.amd.com/rocm/whl/{arch}/ as the Windows ROCm wheel
source. These wheels bundle their own ROCm runtime, support all Python
versions (not just cp312), and include the full torch._C extension set
(including _distributed_c10d) that the old repo.radeon.com wheel omitted.

Changes:
- install.ps1: remove Select-ROCmWheelRelease + hardcoded cp312 wheel
  URLs; remove Python 3.12 forced-preference logic; install via
  --index-url repo.amd.com/rocm/whl/{arch-family}/
- studio/setup.ps1: same -- remove Select-ROCmWheelRelease, switch to
  repo.amd.com arch-aware index URL
- studio/install_python_stack.py: replace _ROCM_WINDOWS_RELEASES /
  _select_windows_rocm_release with _windows_rocm_index_url() using the
  _GFX_TO_AMD_INDEX_ARCH map; drop Python 3.12 restriction
- studio/backend/core/training/worker.py: remove all stub machinery
  (_make_mod_stub, _StubSubpackageFinder, _StubSubpackageLoader,
  _StubClassMeta, torchao/fsdp/dtensor stubs, _c10d_functional ops
  stubs, BNB DLL detection) -- no longer needed with new wheel source

* fix(rocm/win): restore _distributed_c10d + torchao stubs; fix BNB install

repo.amd.com torch wheels also omit torch._C._distributed_c10d on Windows
(RCCL is not shipped on Windows). torch/distributed/__init__.py imports
from it unconditionally at module level, so the stub must land in
sys.modules before any torch.distributed import.

torchao (pulled in by transformers.quantizers) walks
torchao.float8.distributed_utils -> torch.distributed._functional_collectives
-> distributed_c10d at import time. Stubbing torchao up-front short-circuits
that chain.

worker.py:
- Restore _make_mod_stub / _StubSubpackageFinder / _StubSubpackageLoader
- Restore _StubClassMeta for ProcessGroup.BackendType attribute access
- Restore _distributed_c10d stub with __getattr__ (Windows only)
- Restore torchao stubs (5 modules, Windows only)

install_python_stack.py:
- BNB AMD wheel install was inside the early-return branch that fires when
  torch is already a ROCm build (installed by install.ps1). Move BNB install
  outside that branch so it always runs on Windows ROCm — the PyPI
  bitsandbytes has only CUDA DLLs and fails to load on ROCm.

* worker: remove _distributed_c10d stub; stub only torchao

The installed torch/distributed/__init__.py from repo.amd.com
(torch==2.10.0+rocm7.12.0) is now properly guarded with
`if is_available():`, so `import torch.distributed` alone is safe.

The crash only comes via torchao's import chain:
  torchao.float8.distributed_utils
    → torch.distributed._functional_collectives (unguarded import)
    → torch.distributed.distributed_c10d
    → torch._C._distributed_c10d  ← absent on Windows ROCm

Stubbing torchao short-circuits the chain entirely. No need to stub
_distributed_c10d. Remove _StubClassMeta and the _c10d stub block;
keep only _make_mod_stub + _StubSubpackageFinder + torchao seeds.

* fix: BNB AMD wheel skipped + torch.compile segfault on Windows ROCm

install_python_stack.py: the UNSLOTH_ROCM_TORCH_INSTALLED=1 early-return
path (set by setup.ps1 when it installed torch itself) returned before
ever reaching the AMD BNB prerelease wheel install.  The PyPI
bitsandbytes==0.49.x ships only CUDA DLLs, so loading it on ROCm fails
with "libbitsandbytes_rocm72.dll not found".  Now installs the AMD
Windows BNB wheel before returning on that path too.

worker.py: torch._grouped_mm crashes on gfx1200 (null HIP kernel pointer,
0xC0000005) when torch.compile's JitDecomp system dispatches it during
the first forward pass.  Detect Windows ROCm via torch.version.hip
(already in sys.modules from section 1e) and set TORCHDYNAMO_DISABLE=1
to bypass the broken kernel dispatch.

* fix: BNB AMD wheel install fails uv wheel filename check

The bitsandbytes continuous-release wheel is intentionally mismatched:
filename encodes 1.33.7.preview (= 1.33.7rc0 in PEP 440) but wheel
metadata reports 0.50.0.dev0.  uv rejects this by default.

Introduce _install_bnb_windows_rocm() helper that sets
UV_SKIP_WHEEL_FILENAME_CHECK=1 only for this specific install, then
restores the previous env value.  Both BNB install call sites (the
UNSLOTH_ROCM_TORCH_INSTALLED early-return path and the normal Windows
ROCm path) now use this helper.

* worker: patch _grouped_mm CUDA dispatch on Windows ROCm (gfx1200 null kernel)

TORCHDYNAMO_DISABLE=1 stopped the compiler frontend but not the autograd
JitDecomp system, which also dispatches _grouped_mm and hits the same
null HIP kernel crash (0xC0000005).

Verified that torch.library.Library("aten","IMPL").impl("_grouped_mm", fn,
"CUDA") successfully overrides the broken HIP kernel with a Python mm
fallback on torch==2.10.0+rocm7.12.0.

Schema: _grouped_mm(Tensor self, Tensor mat2, Tensor? offs=None,
                    Tensor? bias=None, ScalarType? out_dtype=None) -> Tensor

The fallback handles both the simple case (offs=None → torch.mm) and the
grouped case (offs provided → split self by offsets, multiply each group
against the corresponding slice of mat2, then cat results).

Keep _WINDOWS_ROCM_GROUPED_MM_LIB alive at function scope to prevent the
C++ dispatch registration from being freed by GC.

* worker: fix torchao stub — return stub classes not modules for isinstance()

peft/tuners/lora/torchao.py does:
  from torchao.dtypes import AffineQuantizedTensor, LinearActivationQuantizedTensor
  isinstance(weight, (AffineQuantizedTensor, LinearActivationQuantizedTensor))

The stub __getattr__ was returning stub modules, which isinstance() rejects
with "arg 2 must be a type, a tuple of types, or a union".

Add _StubTypeMeta metaclass whose __instancecheck__ always returns False,
and _make_stub_type() to create stub classes via it. Change _make_mod_stub
__getattr__ to return stub classes instead of stub modules for leaf
attribute access, so isinstance() gets a valid type and returns False.

_StubSubpackageFinder still handles import-style subpackage creation
(those still need module objects in sys.modules); __getattr__ only fires
for from-import or direct attribute access, which are the isinstance paths.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* tests: add coverage for Windows ROCm install paths and worker patches

Add conftest.py to fix pre-existing sys.path issue that prevented
test_rocm_support.py from running at all (install_python_stack.py
imports from backend.utils.wheel_utils which needs studio/ on sys.path).

New test classes cover everything added in this session:
- TestWindowsRocmIndexUrl: arch → AMD pip index URL mapping (gfx120X-all,
  gfx1151, gfx1150, gfx110X-all, unknown → None, trailing slash)
- TestDetectWindowsGfxArch: hipinfo output parsing, missing/timeout/bad
  returncode/no-gcnArchName paths
- TestInstallBnbWindowsRocm: UV_SKIP_WHEEL_FILENAME_CHECK set+restored,
  env restored on exception, no-op when URL missing
- TestRocmTorchInstalledEnvVar: UNSLOTH_ROCM_TORCH_INSTALLED=1 skips
  pip_install, calls _install_bnb_windows_rocm, sets flag
- TestWorkerWindowsRocmPatches: _grouped_mm CUDA dispatch override,
  offs/grouped variant handling, GC-prevention sentinel,
  _StubTypeMeta __instancecheck__, _StubSubpackageFinder registration,
  torchao key submodule pre-stubbing, TORCHDYNAMO_DISABLE guard
- TestRocmTorchPkgSpecs: rocm7.2 torch 2.11.x spec, default <2.11 cap,
  3-tuple shape, _GFX_TO_AMD_INDEX_ARCH RDNA4/3.5/3 coverage

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* tests: fix encoding, IS_WINDOWS patching, and wrong assertion

- Add encoding="utf-8" to all read_text() calls (54 occurrences) so
  tests pass on Windows where the default codec is cp1252 and source
  files contain UTF-8 emoji (e.g. ⚠️ in install_python_stack.py)
- Add @patch.object(stack_mod, "IS_WINDOWS", False) to Linux-path
  TestEnsureRocmTorch tests so they reach the Linux code path when run
  on a Windows machine instead of short-circuiting into the Windows branch
- Fix test_grouped_mm_patch_guarded_by_windows_and_hip_check: the source
  uses getattr(_torch_for_rocm, "version", None) not torch.version, so
  check for '"version"' and '"hip"' substrings instead

137 passed, 2 skipped

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: pin BNB_ROCM_VERSION=72 for torch==2.11.0+rocm7.13.0 compatibility

AMD's pip index now ships torch==2.11.0+rocm7.13.0 (ROCm 7.13).
bitsandbytes auto-detects HIP 7.13 from torch.version.hip and looks for
libbitsandbytes_rocm713.dll, which the AMD Windows prerelease wheel does
not ship (it only ships rocm72.dll), causing a load error at training start.

Fix:
- worker.py section 1f: set BNB_ROCM_VERSION=72 (via setdefault) before
  section 2 ML imports, so bitsandbytes always loads rocm72.dll on Windows ROCm
- install_python_stack.py: set BNB_ROCM_VERSION=72 in _install_bnb_windows_rocm()
  for any post-install imports; update comment to document root cause
- tests: 4 new assertions covering the fix (141 passed, 2 skipped)

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: detect BNB ROCm DLL suffix dynamically instead of hardcoding '72'

BNB_ROCM_VERSION was pinned to '72' which works today (AMD wheel ships
rocm72.dll) but would break again if AMD ships a future wheel with a
different DLL suffix (e.g. rocm713.dll).

Add _detect_bnb_rocm_dll_ver() to install_python_stack.py: scans the
installed bitsandbytes package dir for libbitsandbytes_rocm{VER}.dll
using importlib.util.find_spec (no BNB import needed) and returns the
suffix.  '72' remains the fallback when detection fails.

Apply the same detection inline in worker.py section 1f.  Both paths
still respect a pre-set BNB_ROCM_VERSION (caller override wins).

Tests: +8 cases covering detection logic and fallback (147 passed, 2 skipped).

* fix: patch torch.distributed stubs in server process for Windows ROCm

On Windows ROCm, torch.distributed ships without process-group helpers
(is_initialized, is_available, get_rank, get_world_size).  The worker
subprocess already patches these in section 1e, but the main server
process calls _determine_attention_impl_for_gpu_estimate() which calls
unsloth's resolve_attention_implementation() → is_initialized(), causing:

  "Could not resolve attention implementation for '...':
   module 'torch.distributed' has no attribute 'is_initialized'"

Fix: patch the missing attrs onto torch.distributed at the top of
_determine_attention_impl_for_gpu_estimate, matching the same stubs
already applied in worker.py section 1e.  No-ops on Linux/CUDA where
torch.distributed is fully populated.

* fix: gate _grouped_mm dispatch patch on HIP < 7.13

AMD fixed the gfx1200 null HIP kernel in ROCm 7.13 (torch 2.11+).
Users on the new wheel now get the real GPU _grouped_mm kernel for
MoE workloads instead of the Python mm fallback.

Changes:
- worker.py: add _hip_ver_at_least() helper; wrap full _grouped_mm
  patch in `if not _hip_ver_at_least(7, 13):` with else branch that
  logs the skip reason; update section-1f comment to document the fix
- test_rocm_support.py: add 5 tests covering the helper definition,
  the (7, 13) gate expression, the else branch, the skip log message,
  and the AMD-format version string parsing (.split(".")[:2])

Verified: torch==2.11.0+rocm7.13.0 — 3D batch and grouped (offs)
variants both succeed; null crash only present on rocm7.12 and earlier.

* fix: stub is_torchelastic_launched on torch.distributed for Windows ROCm

resolve_attention_implementation calls is_torchelastic_launched() which
does not exist in the incomplete torch.distributed shipped with the
Windows ROCm wheel, causing a warning on every model config load in the
server process. Add it to the stub table alongside the four helpers
already patched in _determine_attention_impl_for_gpu_estimate.

Also adds two tests: one confirming the new stub and one confirming all
five core distributed helpers are covered.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: explicit warnings on AMD ROCm arch/version fallbacks + Fast-Install arg order

setup.ps1:
- Fix Fast-Install argument order: packages before flags, consistent with
  all other Fast-Install calls in the file
  (was: Fast-Install --force-reinstall --index-url $url torch ...)
  (now: Fast-Install torch torchvision torchaudio --force-reinstall --index-url $url)
- Add explicit [WARN] substep when $HasROCm is true but arch mapping fails:
  - GPU arch detected but not in supported wheel list → names the arch and
    lists supported families so user knows exactly what to report
  - HIP SDK present (amd-smi path) but gcnArchName unreadable → instructs
    user to re-install the HIP SDK; previously fell back silently to CPU

install.sh:
- Add [WARN] to stderr before silent CPU fallback when AMD GPU is confirmed
  (rocminfo/amd-smi) but ROCm version cannot be read from any source
  (amd-smi, /opt/rocm/.info/version, hipconfig, dpkg, rpm)
- Add [WARN] to stderr when ROCm version is too old (< 6.0) with upgrade link

install.ps1 and setup.sh: no changes needed (already handle these paths correctly)

* fix: robust gfx arch detection for Strix Halo / HIP-runtime-only installs

Covers users who have the HIP runtime (amd-smi available) but not the
full HIP SDK (no hipinfo), which is common on Strix Halo iGPU systems.
Without this, $ROCmGfxArch stays null and the installer silently falls
back to CPU-only PyTorch despite a working GPU.

Detection waterfall (setup.ps1 + install.ps1):
  1. hipinfo gcnArchName          -- full HIP SDK (existing, unchanged)
  2. amd-smi list gfx pattern     -- newer amd-smi versions embed arch
  3. amd-smi static --asic        -- ROCm 6+ ASIC details with GFX target
  4. UNSLOTH_ROCM_GFX_ARCH env    -- manual override escape hatch
  5. GPU name → arch table        -- best-effort from marketing name:
       890M / Strix Halo  → gfx1151 (RDNA 3.5 iGPU, Strix Halo)
       880M / Strix Point → gfx1150 (RDNA 3.5 iGPU, Strix Point)
       780M / Phoenix     → gfx1103 (RDNA 3 iGPU)
       RX 7900/7800/7700  → gfx1100 (RDNA 3 desktop)
       RX 9070 XT / 9080  → gfx1201 (RDNA 4)
       RX 9070 / 9060 XT  → gfx1200 (RDNA 4)

When arch is inferred from name, a Cyan substep tells the user to set
UNSLOTH_ROCM_GFX_ARCH to skip inference on future installs.
WMI block intentionally does not set $HasROCm (no runtime confirmation).

Tests: 11 new tests in TestStrixHaloGfxArchDetection covering all five
detection levels, WMI safety, and gfx regex in both ps1 files.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: resolve hipinfo/hipconfig via HIP_PATH/ROCM_PATH when not on PATH

AMD HIP SDK sets HIP_PATH on Windows but does not always add the bin
directory to PATH.  Get-Command hipinfo therefore silently fails and
detection falls through to WMI, which cannot provide a gfx arch, leaving
the user with a CPU-only PyTorch install and no warning.

Changes:
- setup.ps1 / install.ps1: before falling through to amd-smi, attempt to
  locate hipinfo.exe and hipconfig.exe under $env:HIP_PATH\bin (then
  $env:ROCM_PATH\bin) when Get-Command returns nothing
- Emit a [WARN] with the resolved path and a one-liner to permanently fix
  PATH via SetEnvironmentVariable
- Emit a [WARN] when HIP_PATH/ROCM_PATH is set but the exe is still not
  found (incomplete SDK install)
- Emit a [WARN] with the first hipinfo output line when hipinfo runs but
  returns a non-zero exit code (e.g. "no ROCm-capable device detected")
- 18 new tests in TestHipSdkEnvPathResolution; total 183 passed, 2 skipped

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* feat: print HIP SDK path and full hipconfig version in terminal on AMD detection

Both install.ps1 and setup.ps1 now emit substeps under the gpu step when
AMD ROCm is detected:

  gpu  AMD ROCm (gfx1200)
       HIP SDK: C:\Program Files\AMD\ROCm\7.1
       hipconfig: 7.1.51803-d3a86bd04

Previously only the gpu label (e.g. "AMD ROCm (gfx1200)") was shown with
no indication of where the SDK was found or which exact build was active.
The full hipconfig build string (e.g. 7.1.51803-d3a86bd04 instead of just
7.1) is now stored in ROCmVersionFull and also used in setup.ps1's
'rocm' step label.

9 new tests in TestHipSdkDetectedSubstep; total 192 passed, 2 skipped

* fix: Strix rocm7.1 segfault bypass + Ubuntu 24.04 HIP gcc-install-dir

Issue 1 (install.sh): gfx1151/gfx1150 + ROCm 7.1 causes a segfault in
torch._grouped_mm (moe_utils.py:167). The Radeon repo now ships cp313
wheels for rocm-rel-7.1, so _amd_gpu_radeon=true silently lands on the
broken combo. When Strix Halo/Point is detected and TORCH_INDEX_URL is
rocm7.1, override to rocm7.2 PyTorch index, update TORCH_CONSTRAINT, and
set _amd_gpu_radeon=false to bypass the Radeon repo entirely. Emits a
clear [WARN] explaining the segfault and linking to the ROCm upgrade docs.

Issue 2 (setup.sh): ROCm 7.x ships clang-20 which on Ubuntu 24.04+ picks
/usr/lib/gcc/x86_64-linux-gnu/14/ (runtime dir, no C++ headers), causing
'cstdlib file not found' and a failed llama.cpp HIP build. Iterate gcc
versions 14→11 to find the first install dir that has both runtime and
/usr/include/c++/<ver> headers, then pass --gcc-install-dir to clang via
CMAKE_HIP_FLAGS. Fix confirmed by h34v3nzc0dex (llama.cpp 417/417 clean).

11 new tests across TestStrixRocm71Override and TestSetupShGccInstallDir;
total 203 passed, 2 skipped

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: BNB_ROCM_VERSION in server process + torch._C._distributed_c10d stubs

Two errors visible in training logs on Windows ROCm:

1. Server process bitsandbytes crash:
   "Configured ROCm binary not found at libbitsandbytes_rocm713.dll"
   The installed BNB wheel ships rocm72.dll (not rocm713.dll). The
   training worker already sets BNB_ROCM_VERSION=72 via DLL detection
   but the server process (main.py) imported bitsandbytes before that
   ran. Fix: add the same DLL-scan + BNB_ROCM_VERSION assignment to
   main.py inside the existing win32 guard, before any downstream
   import can pull in bitsandbytes.

2. torch.distributed import failure:
   "No module named 'torch._C._distributed_c10d'; torch._C is not a package"
   torch._C is a C extension on Windows ROCm — Python cannot do
   submodule imports from it, so torch.distributed fails to import
   before our attribute stubs could ever run. Fix: inject empty
   ModuleType stubs for _distributed_c10d, _distributed_autograd and
   _distributed_rpc into sys.modules inside the win32 guard in
   hardware.py BEFORE importing torch.distributed, so the import
   succeeds and our attribute stubs take effect.

9 new tests in TestServerStartupRocmFixes; total 212 passed, 2 skipped

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(win32): populate distributed c10d stub with dummy symbols

torch.distributed tries to `from torch._C._distributed_c10d import
FakeProcessGroup` (and ProcessGroup, Work, Store, etc.).  The previous
empty ModuleType stub caused an AttributeError on those names.

Populate every stub with a _Dummy class for each known symbol so the
import chain completes silently on Windows ROCm where torch._C is a
compiled extension and its _distributed_c10d submodule doesn't exist.

Adds four new tests in TestServerStartupRocmFixes covering FakeProcessGroup,
ProcessGroup, setattr population, and all three _distributed_* siblings.

* fix(win32): distinguish HIP SDK installed vs GPU not ROCm-accessible

Previously, when hipinfo was found but exited non-zero (e.g. "no
ROCm-capable device detected"), both install.ps1 and setup.ps1 fell
through to the WMI-label-only branch and printed "AMD GPU detected --
HIP SDK not found" -- factually wrong since the SDK binary is present.

Add $HipSdkInstalled flag (set true when hipinfo binary is found,
regardless of exit code). When HipSdkInstalled && !HasROCm:
- Show "AMD GPU detected -- not ROCm-accessible (HIP <ver>)" instead
- Explain this is a driver issue, not an SDK issue, with a link
- Still run hipconfig version capture so version shows in output
- CPU-only hint now says "GPU not ROCm-accessible" not "require HIP SDK"

Also applies to setup.ps1 (same detection block, same branches).

Adds TestHipSdkInstalledButDeviceInaccessible (11 tests).

* fix(win32): scope ROCm workarounds to AMD hosts only

Three Codex-flagged issues where Windows ROCm workarounds incorrectly
applied to Windows CUDA (NVIDIA) machines:

main.py (P1): BNB_ROCM_VERSION was set unconditionally on all win32
hosts. On NVIDIA, bitsandbytes sees BNB_ROCM_VERSION and looks for a
ROCm DLL that doesn't exist, breaking bitsandbytes initialisation.
Fix: gate the block on HIP_PATH/ROCM_PATH being present (ROCm hosts only).

worker.py (P2): torchao stubs were seeded for all win32 runs, shadowing
real torchao on Windows CUDA and silently disabling torchao quantization
for NVIDIA users. Fix: gate on HIP_PATH/ROCM_PATH (win32 ROCm only).

install_python_stack.py (P1): _detect_windows_gfx_arch() only checked
shutil.which("hipinfo"), skipping the HIP_PATH/ROCM_PATH fallback that
the PowerShell installers use. On installs where the HIP SDK bin dir is
not on PATH, _ensure_rocm_torch() returned early without installing
ROCm wheels or bitsandbytes. Fix: mirror the env-var fallback.

* fix(linux): route Strix + ROCm 7.1 to AMD arch-specific index

Instead of falling back to pytorch.org/rocm7.2, the Strix override now
routes to repo.amd.com/rocm/whl/gfx1151/ (or gfx1150/) which serves
torch 2.11.0+rocm7.13.0 -- AMD's build containing the actual _grouped_mm
kernel fix, verified on real gfx1151 hardware by h34v3nzc0dex.

This exercises the real GPU kernel path rather than the rocm7.2 workaround.
UNSLOTH_AMD_ROCM_MIRROR can override the base URL for air-gapped installs.

Also teaches _tauri_torch_index_family to recognise AMD arch-specific URLs
(repo.amd.com/rocm/whl/gfx*) and return the rocm7.13 family label so
_tauri_gpu_branch correctly classifies these installs as rocm.

Suggested by h34v3nzc0dex based on hardware-verified probe results.

* fix(studio/rocm): gate ROCm-only side-effects on active torch runtime

Address five edge cases flagged during PR review:

1. studio/backend/main.py: BNB_ROCM_VERSION was set whenever HIP_PATH or
   ROCM_PATH was present in the environment. A Windows CUDA user who once
   installed the HIP SDK and reverted to a CUDA torch wheel still has those
   env vars set, so bitsandbytes would try to load libbitsandbytes_rocm72.dll
   against a CUDA torch and crash. Now probe torch.version.hip inside the
   env-var guard (worker.py already does this).

2. studio/backend/main.py: os.add_dll_directory returned handles were
   discarded. Per CPython docs, the directory leaves the DLL search list when
   the handle is garbage collected. Retain handles in module-level
   _ROCM_DLL_HANDLES list so they survive process lifetime.

3. studio/install_python_stack.py: _install_bnb_windows_rocm() returned None
   regardless of pip_install_try outcome, and the caller flipped
   _rocm_windows_torch_installed to True unconditionally. On a failed BNB
   install the post-install "manual install may be required" warning was
   suppressed and the user was misled. Helper now returns bool; caller gates
   on it.

4. studio/install_python_stack.py: _detect_windows_gfx_arch returned the raw
   capture group, so mixed-case hipinfo output ("Gfx1151") missed the
   lowercase keys in _GFX_TO_AMD_INDEX_ARCH and silently fell back to CPU
   torch. Lowercase the token.

5. studio/install_python_stack.py: UNSLOTH_ROCM_TORCH_INSTALLED=1 early-
   return trusted the env var even when the venv was wiped between runs.
   Subprocess-probe torch importability first; fall through to the full
   install path if the probe fails.

Tests: 231 passed, 1 skipped in tests/studio/install/test_rocm_support.py
(adds one new test for case 5 fall-through).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio/rocm): worker.py parity + don't roll back ROCm torch on bnb failure

Addresses findings from a 10x reviewer pass on the prior fix commit:

1. studio/backend/core/training/worker.py (parity with main.py):
   - Gate the torchao stub block on torch.version.hip / 'rocm' in
     torch.__version__ instead of HIP_PATH / ROCM_PATH env-var presence.
     Same root cause as main.py: HIP SDK env vars stick around on CUDA hosts.
   - Add module-level Windows ROCm DLL registration block. Worker subprocesses
     inherit env vars but not the parent's add_dll_directory handles, so the
     first `import torch` in the worker could fail to find amdhip64.dll when
     HIP_PATH\bin is not on PATH. Mirrors main.py setup. Handles retained at
     module scope via _ROCM_DLL_HANDLES.
   - Promote _WINDOWS_ROCM_GROUPED_MM_LIB to module scope with `global` in
     run_training_process so the torch.library.Library registration survives
     past function return / mid-run garbage collection.
   - Harden _torch_has_hip() to also accept 'rocm' in torch.__version__
     (AMD SDK / Radeon wheels may not set torch.version.hip).

2. studio/install_python_stack.py:
   - Don't roll back ROCm torch when bitsandbytes install fails. The prior
     commit gated _rocm_windows_torch_installed on _install_bnb_windows_rocm()
     returning True; if torch installed successfully but bnb failed, the flag
     stayed False and later install steps could overwrite ROCm torch with the
     generic CPU torch wheel. Set the flag after torch install; surface bnb
     failure as a separate warning instead.
   - _detect_windows_gfx_arch now probes in three tiers: UNSLOTH_ROCM_GFX_ARCH
     env-var override (matches the PowerShell installer), then hipinfo (PATH
     or HIP_PATH\bin), then amd-smi (`static --asic`, `list`). Without the
     amd-smi fallback, runtime-only Radeon installs without hipinfo on PATH
     made `studio update` return early and leave the venv on CPU torch.
   - Linux torch-already-rocm probe in _ensure_rocm_torch now matches the
     Windows probe shape: accepts torch.version.hip OR 'rocm' in
     torch.__version__ to cover AMD SDK / Radeon Linux wheels.

3. studio/backend/utils/hardware/hardware.py:
   - apply_gpu_ids() final-fallback torch probe accepts 'rocm' in
     torch.__version__ in addition to torch.version.hip, matching
     detect_hardware(). AMD SDK wheels could otherwise leak through with
     CUDA-only visibility masks on a spawned ROCm worker.

Tests: 231 passed, 1 skipped in tests/studio/install/test_rocm_support.py
(no test changes needed; the probe shape that prints the hip version (or
'rocm' sentinel) preserves the existing non-empty-string contract).

Not addressed in this commit (deferred or out of scope):
- Tag drift / lemonade checksum (PR 5303 surface, not this PR).
- install.sh rocm7.2.1 URL: small fix, separate.
- install.ps1 / setup.ps1 'Radeon 8060S' marketing-name fallback table.
- Strix Halo + ROCm 7.1 routing asymmetry in Python update path.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio/rocm): robustness pass - rocm tag normalisation, Strix routing parity, hardened detection

Robustness pass on top of 76137b2d. Four targeted fixes:

1. install.sh ROCm-tag routing normalisation.
   `rocm7.2.1` would route to https://download.pytorch.org/whl/rocm7.2.1
   which does not exist (PyTorch publishes major.minor URLs only). Same
   for any future patch-level tag. Normalise every rocm{maj.min}* pattern
   to the bare {maj.min} index URL.

2. install.ps1 + studio/setup.ps1 marketing-name fallback.
   The gfx1151 row matched 890M / Strix Halo / HX 37x / HX 38x / AI 9 HX
   but not the actual retail name 'AMD Radeon 8060S Graphics' shipped by
   OEMs (Ryzen AI MAX+ 395). Add '8060S' to the regex.

3. install_python_stack.py Strix + ROCm 7.1 routing parity with install.sh.
   The shell installer reroutes Strix Halo / Point + ROCm 7.1 to
   repo.amd.com/rocm/whl/{gfx}/ (which serves torch 2.11.0+rocm7.13.0
   with the upstream _grouped_mm fix). The Python `studio update` path
   only warned and still installed the broken generic rocm7.1 wheel.
   Mirror the override: detect gfx1151/gfx1150 on ROCm 7.1, route to
   the AMD per-gfx index, honour UNSLOTH_AMD_ROCM_MIRROR override.

4. _detect_windows_gfx_arch amd-smi parsing tightened.
   The amd-smi fallback added in the prior commit used a bare
   `\bgfx[1-9][0-9a-z]{2,3}\b` match against the lowercased stdout,
   which could pick up stray gfx references in warnings / device-name
   strings. Anchor on labelled lines first (Target_Graphics_Version,
   ASIC, Arch, gfx) and fall back to the bare match only when no
   labelled line is present.

Tests: 231 passed, 1 skipped in tests/studio/install/test_rocm_support.py;
sim_5301 23 cases pass (6 new sims for the Strix override + amd-smi parsing).

* fix(studio/rocm): multi-GPU selection, Strix sibling handling, defensive cleanups

Round 4 robustness pass based on 5 parallel Opus reviewers of head 21773215.
Seven items from across regression / edge-case / error-paths / architecture
reviews:

1. studio/backend/main.py BNB gate: aligned with the broad ROCm check used
   everywhere else in this PR (torch.version.hip OR 'rocm' in __version__).
   AMD SDK / Radeon Linux wheels do not always populate torch.version.hip;
   without this, main.py would silently skip BNB_ROCM_VERSION while worker.py
   set it.

2. studio/install_python_stack.py _install_bnb_windows_rocm: init _ok = False
   before the try block. Without this, if pip_install_try itself raises
   (e.g. OSError on uv binary missing), the finally block restored env vars
   correctly but the subsequent `if not _ok:` raised UnboundLocalError,
   masking the original exception.

3. studio/install_python_stack.py _detect_windows_gfx_arch:
   - Rewrote to use re.findall (not re.search) on both hipinfo and amd-smi
     output, dedup tokens preserving order, and select via new
     _pick_visible_index() helper.
   - HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES (first comma entry, integer)
     now picks the right GPU on multi-AMD-GPU hosts. Out-of-range or non-int
     values fall back to the first GPU (matches detect_host behaviour in
     install_llama_prebuilt.py).

4. studio/install_python_stack.py Strix override now consults the runtime
   target before flipping:
   - Previous behaviour intersected gfx_codes with {gfx1151, gfx1150} and
     picked the first Strix arch, ignoring whether HIP_VISIBLE_DEVICES
     selected a non-Strix sibling (e.g. discrete RX 7900 in a mixed APU+dGPU
     box). Could install Strix-specific wheels onto a gfx1100 dGPU.
   - Now resolves the runtime gfx via _pick_visible_index() and only
     overrides when that runtime target is in the Strix set.

5. studio/backend/main.py + studio/backend/core/training/worker.py: ROCm
   version dir scan no longer sorts lexically. Previous sort placed "10.0"
   before "7.0" alphabetically, which would mis-prioritise ROCm 10.x bin
   dirs once AMD ships them. New _ver_key() splits on "." and sorts
   numerically with a string fallback.

6. install.sh Strix override URL: replaced ${var%/} (strips one trailing
   slash) with a while-loop that strips all trailing slashes, matching
   Python's .rstrip("/"). A user setting UNSLOTH_AMD_ROCM_MIRROR with
   "http://corp/whl///" no longer ends up with "http://corp/whl///gfx1151/"
   which strict pip proxies (artifactory, sonatype) 404 on.

7. studio/install_python_stack.py: bumped torch import probe timeout from
   30s to 90s. PyTorch's lazy .so loading can take 60-90s on cold NFS or
   USB-backed venvs. The shorter timeout was producing a false "torch
   missing" classification and reinstalling a working ROCm torch.

Tests: 231 passed, 1 skipped. sim_5301 30 cases pass (added 7 new sims for
multi-GPU detection, Strix sibling handling, and _ok-init regression).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio/rocm): worker BNB/grouped_mm broad gate, install.sh Strix visibility, runtime-only ROCm detection

Round-5 robustness pass based on 20 parallel reviewers of head 96b9e465.

1. studio/backend/core/training/worker.py - BNB version pin / dynamo disable
   / _grouped_mm fallback block was still gated on torch.version.hip alone
   despite the torchao stub block above already using the broad check. AMD
   SDK / Radeon Windows wheels (torch.__version__ contains "rocm" but
   torch.version.hip is None) silently skipped the Windows ROCm runtime
   patches. Aligned to the same broad check (8/20 reviewers).

2. studio/backend/core/training/worker.py - _hip_ver_at_least() now also
   parses the ROCm version out of torch.__version__ (e.g. "2.11.0+rocm7.13.0")
   when torch.version.hip is missing, so the kernel-fix gate is correct for
   SDK / Radeon wheels too.

3. studio/backend/core/training/worker.py - _grouped_mm_safe_impl with
   offs=None now picks torch.bmm/matmul for 3-D inputs instead of always
   calling torch.mm. The real _grouped_mm accepts 3-D batched matmul; the
   prior fallback raised "self must be a matrix" on MoE workloads (2/20).

4. studio/backend/main.py - dropped the HIP_PATH / ROCM_PATH env-var gate
   from the BNB block; probe torch directly. Runtime-only Radeon / AMD SDK
   Windows installs do not set those SDK env vars but still ship ROCm torch
   (5/20 reviewers).

5. install.sh - Strix override now collects every gfx token from
   rocminfo / amd-smi (in enumeration order), then indexes by
   HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES so a mixed Strix iGPU + non-
   Strix dGPU host where the user selected the dGPU does NOT get rerouted
   to the Strix per-gfx index. Mirrors the Python update path (5/20 reviewers).

6. install.sh - Strix detection chain now also probes `amd-smi static --asic`,
   matching the PowerShell installer (1/20). Closes the gap on runtime-only
   Strix hosts where `amd-smi list` does not surface a gfx token.

7. studio/install_python_stack.py - _has_rocm_gpu() now has the sysfs KFD
   topology fallback (/sys/class/kfd/kfd/topology/nodes/*/gpu_id), matching
   install.sh. On minimal package-managed installs without rocminfo /
   amd-smi GUI tools, `studio update` can now detect the GPU and repair the
   venv instead of returning early (2/20).

8. studio/install_python_stack.py - _detect_amd_gfx_codes() now falls back
   to `amd-smi list` and `amd-smi static --asic` when rocminfo is missing
   (2/20). Strix routing on runtime-only Radeon hosts now matches what
   install.sh has done for a while.

9. studio/install_python_stack.py - Strix override now applies even when
   has_hip_torch is True. The whole point of the override is to repair an
   existing broken torch.version.hip == "7.1" install; skipping the
   reinstall left users on the known _grouped_mm segfaulting stack (3/20).

Tests: 231 passed, 1 skipped. sim_5301 30 cases pass. sim_cross 12 pass.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio/rocm): code review hardening pass

- main.py: numeric DLL sort (string sort picked rocm72 over rocm713);
  add basename() to regex; log warning on detection failure; log info
  when BNB_ROCM_VERSION is set (mirrors worker.py)
- worker.py: explicit len-guard in _hip_ver_at_least() with warning
  logs instead of silent IndexError/ValueError swallow
- hardware.py: isinstance(result, dict) guard before result.get() in
  _smi_query() to prevent AttributeError on non-dict backend returns
- amd.py: round() before int() on parsed GPU IDs; log warning when
  truncation occurs (defensive against malformed amd-smi output)
- setup.sh: quote --gcc-install-dir value in CMAKE_HIP_FLAGS so paths
  with spaces do not break the CMake argument
- install.ps1, setup.ps1: apply colon-split + ToLower() to hipinfo
  gcnArchName match (consistent with each other and with setup.sh)
- install.sh: tighten ROCm tag case patterns to explicit
  rocmX.Y|rocmX.Y.* to avoid unintended prefix matches

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio/training): GPU OOM guard to prevent system freeze on VRAM exhaustion

On RDNA 4 (gfx1200/gfx1201) and other ROCm GPUs, exhausting VRAM can
cause a HIP driver hang that freezes the entire system rather than
raising a recoverable Python exception.

Two-part fix:
- set_per_process_memory_fraction(0.90) caps the HIP/CUDA allocator at
  90% of VRAM so PyTorch raises OutOfMemoryError before hitting the
  hardware limit, keeping the driver alive and the system responsive
- top-level exception handler detects OOM errors by type and message
  and surfaces a clear actionable message to the UI (reduce
  max_seq_length, enable gradient_checkpointing, lower batch size)
  instead of the raw CUDA/HIP error string

* fix(studio/rocm): OOM guard ROCm-only + unified memory, multi-GPU arch selection

OOM guard (worker.py):
- Scope to _hw.IS_ROCM only -- NVIDIA CUDA has a graceful OOM path and
  does not need the allocator cap
- Detect unified memory by comparing torch VRAM against psutil system RAM;
  use 0.80 on unified-memory APUs (gfx1151 Strix Halo) where the GPU pool
  is carved from host RAM, 0.90 on discrete cards

Multi-GPU arch selection:
- install.ps1 / setup.ps1: replace -match (first hit only) with
  [regex]::Matches() to collect all gcnArchName entries, then index by
  HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES
- install_python_stack.py: index into full token list before dedup so
  HIP_VISIBLE_DEVICES=2 on [gfx1100, gfx1100, gfx1151] resolves gfx1151
- install.sh: remove awk dedup from gfx token collection for same reason

GCC multiarch (setup.sh):
- Only append -linux-gnu when gcc -print-multiarch does not already return
  the full triple, fixing double-suffix on Ubuntu 24.04

* fix(tests): update ROCm version cap expectations from rocm7.1 to rocm7.2

Daniel's normalisation commit updated the cap from rocm7.1 to rocm7.2
since PyTorch now publishes that index and rocm7.2 ships torch 2.11.0.
Test expectations were stale.

* fix(tests): correct MLX smoke test losses_per_step assertion

logging_steps=1 with max_steps=30 produces 30 loss entries, not 7.
The assertion was stale from a previous config.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(studio/worker): detect unified-memory APU by GPU name not VRAM/RAM ratio

The previous heuristic (VRAM > 50 % of system RAM) false-positived on discrete
cards in low-RAM systems — e.g. RX 9060 XT 16 GB on a 16 GB or 24 GB machine
would trip the unified-memory path and log "unified memory host" when it should
say "discrete".

AMD iGPUs (gfx1150/gfx1151 Strix Halo, Strix Point, etc.) expose names with a
digit+M suffix ("AMD Radeon 890M"), while discrete cards use "RX NNNN [XT|XTX]"
naming.  Matching that suffix is reliable across all current ROCm-capable AMD
consumer GPUs and does not require psutil.

Also includes the device name in the log line to ease future debugging.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(install/setup.ps1): force array on hipinfo gcnArchName parse to fix single-GPU arch truncation

When [regex]::Matches() finds exactly one match, PowerShell's pipeline
unwraps the result to a scalar string.  Indexing a scalar string with [0]
returns the first *character*, so a one-GPU system would parse
gcnArchName "gfx1200" as "g", which is not in the supported arch map
and triggers the CPU-only fallback.

Wrapping with @() forces the result to remain an array regardless of
match count.  On a single-GPU machine the arch is now correctly read as
"gfx1200" (or whatever the full name is) so the ROCm wheel index is
selected.

Reproducer: hipinfo exits 0 and outputs exactly one gcnArchName line.
Without @(), $_hipAllArches = "gfx1200" (String); $_hipAllArches[0] = 'g'.
With @(), $_hipAllArches = @("gfx1200") (Object[]); $_hipAllArches[0] = "gfx1200".

* fix(studio/rocm): classify unified-memory APU via VRAM/RAM ratio, not arch list

Replace the gcnArchName allowlist {gfx1150, gfx1151} with a
psutil-based heuristic: unified APUs expose the entire system RAM
as the HIP pool (ratio ≥ 0.90), discrete cards are well below that.
No arch name required — future APUs classify correctly without code changes.

Also removes the stale import re / \d[Mm]\b device-name regex that
5d84704 left behind, and logs vram/sys GiB for easier on-hardware
verification.

Addresses h34v3nzc0dex review: Radeon 8060S (gfx1151, 128 GiB
unified) now correctly gets 0.80 cap instead of 0.90.

* fix(studio/rocm): revert to gcnArchName for unified-memory APU classification

VRAM/RAM ratio >= 0.90 false-positives on machines where discrete VRAM
equals system RAM (e.g. RX 9060 XT 16 GB + 16 GB system RAM → ratio 1.0,
incorrectly classified as unified → wrong 0.80 cap applied).

gcnArchName is the correct signal: naming-independent, stable within a
product family, and already parsed throughout this PR. Unified set is
{gfx1150, gfx1151} (Strix Point + Strix Halo).

* fix(studio/llama-prebuilt): resolve hipinfo via HIP_PATH/ROCM_PATH on Windows

shutil.which("hipinfo") returns None when the HIP SDK bin dir is not on
PATH -- the HIP SDK installer sets HIP_PATH/ROCM_PATH but does not always
add the bin dir to PATH. This caused has_rocm=False in the prebuilt asset
selector, so AMD ROCm machines got the CPU llama.cpp zip instead of the
HIP one, silently running all chat inference on CPU.

Add _resolve_exe() that falls back to %HIP_PATH%\bin and %ROCM_PATH%\bin
when shutil.which() finds nothing, mirroring the same fallback already
present in setup.ps1.

* fix(studio/llama-prebuilt): pass --has-rocm from setup.ps1 to skip re-detection

The Python prebuilt installer re-detects ROCm independently via
shutil.which("hipinfo"), which fails when hipinfo is not on PATH
(HIP SDK sets HIP_PATH but doesn't always add the bin dir to PATH).
This caused has_rocm=False and downloaded the CPU llama.cpp zip even
on confirmed AMD ROCm machines.

setup.ps1 already performs reliable ROCm detection with its own
HIP_PATH/ROCM_PATH fallback. Add --has-rocm flag to
install_llama_prebuilt.py so setup.ps1 can forward its result directly,
and pass it whenever $HasROCm is true. The Python script then overrides
has_rocm=True in the HostInfo without re-probing.

* fix(studio/llama-prebuilt): add HIP asset to simple-policy Windows path

direct_upstream_release_plan (used by --simple-policy, which setup.ps1
always passes) only checked has_usable_nvidia on Windows and fell
straight to CPU for AMD ROCm machines, ignoring has_rocm entirely.
The --has-rocm override had no effect because the simple-policy code
path never reached resolve_asset_choice where has_rocm was checked.

Add an elif branch for has_rocm that tries the upstream HIP asset
(llama-TAG-bin-win-hip-radeon-x64.zip) before falling through to the
CPU fallback, consistent with the non-simple-policy path.

* fix(studio/setup.ps1): auto-remove mismatched llama.cpp install kind

When an existing llama.cpp install is the wrong kind for the current
GPU (e.g. windows-cpu on an AMD ROCm machine that should have
windows-hip), the prebuilt installer skips on tag match and never
upgrades. Read install_kind from UNSLOTH_PREBUILT_INFO.json before
invoking the installer and remove the directory if the kind doesn't
match, forcing a fresh download of the correct variant.

* fix(studio/setup.ps1): show live PyTorch install output in verbose mode for ROCm

The ROCm torch reinstall (setup.ps1 phase) always silently captured
output, so in --verbose mode the torch downgrade mid-install
(2.11.0+rocm → 2.10.0 → 2.11.0+rocm) looked like the final state was
2.10.0. Match the CPU/CUDA blocks which show live uv output when
$script:UnslothVerbose is set.

* fix(rocm/windows): set ROCBLAS_TENSILE_LIBPATH for bundled rocblas.dll

The llama.cpp ROCm prebuilt bundles rocblas.dll next to the binary but
not the Tensile kernel library files it depends on at runtime
(rocblas/library/TensileLibrary*.dat + *.hsaco).  The bundled DLL
searches for these files relative to its own location by default, i.e.
<binary_dir>/rocblas/library/, which does not exist in the prebuilt
install tree.  This causes a silent crash on the very first GEMM
(prefill) with no output from llama-server, seen by the caller as
WinError 10054 / 10061.  Model load and the single-token warmup pass
because they use simpler code paths that do not trigger rocBLAS GEMM.

Fix: set ROCBLAS_TENSILE_LIBPATH in the subprocess env to
<HIP_PATH>/bin/rocblas/library so the bundled DLL finds the kernel
files from the system ROCm installation.  Uses setdefault so a user-
supplied env var is never overwritten.  No-ops on CUDA and CPU (no
HIP_PATH) and on Linux (win32 branch only).

Reproducer log:
  rocBLAS error: Cannot read .../Release/rocblas/library/TensileLibrary.dat
  rocBLAS error: Could not initialize Tensile host:
  directory_iterator: The system cannot find the path specified.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(install.sh): restore gfx token dedup in Strix multi-GPU awk indexer

536a54df removed the per-source `| awk '!seen[$0]++'` dedup from the
_gfx_all collection step but left the indexer awk as bare NF, so on a
mixed-arch host (e.g. dGPU gfx1100 + Strix iGPU gfx1151) where
rocminfo emits each gfx token twice (Name: field + ISA triple),
HIP_VISIBLE_DEVICES=1 indexed vals[1] = the second gfx1100 occurrence
instead of gfx1151, triggering the Strix routing on the wrong GPU.

Add !seen[$0]++ to the indexer awk so duplicate tokens from the same
GPU collapse to one entry before the HIP_VISIBLE_DEVICES index is
applied -- matching exactly what the Python side does with dict.fromkeys()
in _detect_amd_gfx_codes(). The comment above the block ("skip
duplicates") already documented this as the intended behaviour.

* fix(studio/install): correct _TOTAL progress count on Windows

base_total += 3 fired for all non-macOS platforms including Windows,
but flash-attn (line 1620) and ROCm torch final (line 1705) are both
guarded by 'not IS_WINDOWS and not IS_MACOS', so on Windows with torch
enabled _TOTAL was 13 while only 11 _progress() calls actually execute.

Split into +1 for the ROCm torch check (all non-macOS) and +2 for the
two Linux-only steps, so Windows gets _TOTAL=11 and Linux gets 14.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(install.ps1): enforce torch>=2.11.0 for gfx120X and Strix on Windows

The AMD arch-specific index (repo.amd.com/rocm/whl/gfx120X-all/ and
gfx1151/) publishes torch wheels from 2.7.1 through 2.11.0. Without a
version floor pip can resolve to torch 2.10.0+rocm7.12 on RDNA 4
(gfx120X) or torch 2.10.0+rocm7.1 on Strix (gfx1151/gfx1150), both of
which have a null-pointer crash in torch._C._grouped_mm (TheRock
issues #5284 / #3284). torch 2.11.0+rocm7.13 contains the fix.

Add $ROCmTorchFloor alongside $ROCmIndexUrl: set to torch>=2.11.0 for
the two affected arch families, null for all others. Wire it into the
uv pip install call so the broken wheels are never selected.

* fix(rocm/windows): address Codex nits - deterministic DLL suffix, CUDA llama.cpp kind, HIP_VISIBLE_DEVICES arch indexing

- install_python_stack.py / worker.py: _detect_bnb_rocm_dll_ver() and the
  inline worker probe now collect ALL libbitsandbytes_rocm*.dll suffixes and
  return max() by numeric value instead of stopping at the first glob hit.
  Filesystem glob order is not guaranteed; this ensures '713' always wins
  over '72' when both variants are present in the wheel.

- setup.ps1 (expectedKind): add 'windows-cuda' branch so NVIDIA hosts are
  not treated as 'windows-cpu'. Previously an existing windows-cuda prebuilt
  was always considered a mismatch on non-ROCm machines, forcing an
  unnecessary re-download on every update.

- setup.ps1 (amd-smi gfx arch): collect ALL gfx tokens from amd-smi list
  output in GPU order and honour HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES
  when selecting which arch to use. On mixed-arch AMD systems where the
  visible GPU is not the first enumerated one, this prevents installing an
  incompatible wheel index. Falls back to index 0 (same as before) when the
  visibility var is unset or is a comma-separated list.

- test_rocm_support.py: add test_picks_highest_suffix_when_multiple_dlls to
  cover the multi-DLL case that was previously untested.

* fix(rocm): misleading amd-smi log, BNB spec consistency, torch ceiling for AMD index

amd.py: split 'returncode != 0 or not stdout' into two separate branches.
Previously, exit-0 with empty output logged 'amd-smi returned code 0' (which
reads as success, not a warning) and incorrectly incremented the circuit-breaker
counter. Now: non-zero exit logs the code and counts toward the limit as before;
empty stdout on exit 0 logs at DEBUG level and does not penalise the counter
(amd-smi --json always emits at least [] on exit 0, so this branch is rare and
is not a tool failure).

main.py: replace spec.origin / os.path.dirname() with
spec.submodule_search_locations to match install_python_stack.py and worker.py.
For normal wheel installs both approaches reach the same directory, but using
submodule_search_locations is the canonical way and handles editable bitsandbytes
installs correctly. Also use max() by numeric suffix (same as the other two sites)
instead of a sort-then-break loop.

install.ps1: add <2.12.0 ceiling to the torch constraint for gfx120X (RDNA 4)
and gfx1151/gfx1150 (Strix). AMD actively publishes new versions on their
per-arch index; without a ceiling, a future 2.12.0+rocmX.Y wheel would be
pulled in automatically before being validated on these architectures. The
ceiling matches the existing Linux install_python_stack.py constraint for the
same arches. Bump both when 2.12.x is confirmed working.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(rocm): torch floor in setup.ps1, torchvision pin for Strix, rocmsdk in _hip_ver_at_least

setup.ps1: add \ (mirrors install.ps1) and derive \
from it. Previously the AMD index install called 'Fast-Install torch torchvision
torchaudio --force-reinstall --index-url \' with no version
constraint, so pip could resolve torch 2.10.0+rocm7.12 for gfx1151/gfx1200 --
the exact broken wheel the PR is meant to avoid. Now gfx120X and Strix enforce
'torch>=2.11.0,<2.12.0', matching install.ps1 and the Linux constraint.

install_python_stack.py: pin torchvision and torchaudio in _strix_override_pkgs.
The Strix Linux override uses --index-url (exclusive, no PyPI fallback); bare
unversioned 'torchvision' and 'torchaudio' could resolve a build from AMD's
index targeting a different torch major, causing ABI/version mismatches at
runtime. Now pinned to '>=0.26.0,<0.27.0' and '>=2.11.0,<2.12.0' respectively,
matching _ROCM_TORCH_CONSTRAINT['rocm7.2'].

worker.py: extend _hip_ver_at_least to handle AMD SDK wheel version strings.
The fallback regex r'rocm(\d+)\.(\d+)' cannot match '2.9.0+rocmsdk20251116'
(no rocmX.Y component), so the function always returned False on SDK/Radeon
wheels -- installing the Python _grouped_mm workaround on wheels that already
have the working HIP kernel. Added a second check: if the version string
contains '+rocmsdk', assume >= 7.13 (the rocmsdk format post-dates the
gfx120X null-kernel fix) and skip the fallback.

* fix(rocm): warn on OOB HIP_VISIBLE_DEVICES, bail on empty numeric_ids mask

- setup.ps1: when HIP/ROCR_VISIBLE_DEVICES names an index beyond the
  detected GPU count, emit a yellow warning and fall back to GPU 0
  instead of silently reading allGfxArches[-1] (wrong arch)
- hardware.py _reconcile_primary_rocm_unified_memory: distinguish
  numeric_ids=None (no env var, use torch ordinal 0) from numeric_ids=[]
  (empty mask / HIP_VISIBLE_DEVICES=-1, no GPU visible); bail out early
  in the empty case to avoid querying torch.device(0) incorrectly

* fix(rocm): gate StubSubpackageFinder on win32 ROCm, add gcnArchName fallbacks

- worker.py _StubSubpackageFinder: the meta_path append was running on
  every platform on every call to run_training_process; moved it inside
  the if _is_win32_rocm: block since stubs are only seeded there and the
  finder is a pure accumulation on Linux/Windows CUDA
- worker.py OOM guard: AMD SDK / Radeon wheels may not populate
  gcnArchName, causing Strix Halo to be misclassified as discrete and
  get the 0.90 cap (12.8 GB OS headroom) instead of 0.80 (25.6 GB);
  now tries gcn_arch_name / arch_name / gfx_arch_name variants first,
  then falls back to device-name matching (890M -> Strix Halo,
  880M -> Strix Point) with a debug log when the fallback fires

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(rocm): pin torchvision/torchaudio in setup.ps1, remove -Unique from arch array

- setup.ps1 ROCm torch install: torchvision and torchaudio were passed
  bare alongside pinned torch>=2.11.0,<2.12.0 for gfx1151/gfx1200 arches.
  AMD publishes packages independently so a future torchvision 0.27 (for
  torch 2.12) on the same arch index would cause pip ResolutionImpossible
  or an ABI-incompatible install. Added torchvisionFloorMap and
  torchaudioFloorMap mirroring install_python_stack.py's strix override
  (torchvision>=0.26.0,<0.27.0, torchaudio>=2.11.0,<2.12.0) and derived
  ROCmVisionSpec/ROCmAudioSpec used in all three Fast-Install call sites.

- setup.ps1 amd-smi arch detection: Select-Object -Unique was collapsing
  same-arch multi-GPU arrays (e.g. two gfx1151 APUs -> 1-element array)
  causing HIP_VISIBLE_DEVICES=1 to trigger a false out-of-range warning
  and fall back to GPU 0 even though the correct GPU would have been at
  index 1. Removed -Unique; added comment noting the positional-index
  assumption and its non-contiguous-GPU limitation.

* fix(rocm): add 8060s/8050s to OOM guard device-name fallback, extract classifier helper

Path 3 of the OOM guard device-name fallback only checked for 890m/880m
(gfx1150 Strix Point SKU names). Strix Halo (gfx1151) ships as Radeon 8060S
(Ryzen AI MAX+ 395) and Radeon 8050S (cut-down SKU) -- neither matches, so
the fallback returned is_unified=False and applied the 0.90 fraction instead
of 0.80, leaving ~12.8 GiB OS headroom on a 128 GiB pool instead of ~25.6 GiB.

Fix: add 8060s and 8050s to the name-match set. Also correct the comment that
mislabelled 890M as a Strix Halo name (it is Strix Point).

Refactor: extract the three-path classifier into _rocm_classify_unified_memory()
so it can be unit-tested directly. Add 31 test cases in test_rocm_oom_guard.py
covering all three paths and the regression case (Radeon 8060S Graphics).

Reported-by: h34v3nzc0dex

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix(rocm): pass explicit dtype on bf16-unsupported hardware (RDNA2)

dtype=None lets unsloth auto-detect the model dtype. On RDNA2 (gfx103x,
e.g. RX 6600) is_bfloat16_supported() incorrectly returns True, so unsloth
picks bf16 and the first bf16 kernel dispatch triggers:

  LLVM ERROR: Cannot select: intrinsic %llvm.amdgcn.fdot2.bf16.bf16

Replace every dtype=None in load_model() with _auto_dtype which resolves
to None when bf16 is supported (all modern NVIDIA + RDNA3+) and
torch.float16 otherwise. This gives RDNA2 users a working float16
training path without touching NVIDIA behaviour at all.

Fixes: https://github.com/unslothai/unsloth/issues/5337

* fix: reduce log noise for expected non-issues on Windows ROCm

Three log lines fired at warning/error level for conditions that are
completely expected on a Windows HIP SDK-only setup:

amd.py
- amd-smi WinError 2 (FileNotFoundError): downgrade warning -> debug.
  amd-smi ships with Adrenalin, not the HIP SDK; absence is normal.
- 'disabling' message: downgrade warning -> info with clearer text
  'not available (not installed; expected on HIP SDK-only systems);
  GPU VRAM polling disabled'

hardware.py
- torch.distributed.Store missing: downgrade warning -> debug.
  The distributed stub added in this PR intentionally omits Store; the
  attention-impl fallback to eager is expected and non-actionable.

worker.py
- causal-conv1d: add early Windows exit (info) in both
  _ensure_causal_conv1d_fast_path and _causal_conv1d_install hook;
  no cp313/win_amd64 wheel exists, so the install always fails.
- FLA: add early Windows exit (info) in
  _ensure_flash_linear_attention_unconditional; triton dependency has
  no cp313/win_amd64 wheel.
- Defense-in-depth: _install_package_wheel_first non-HIP PyPI failure
  logs info+debug on Windows instead of error; FLA failure logs
  info+debug on Windows instead of warning.

* [AMD] FIx installation of bitsandbytes when it's from .dev and skip rebuilding llama.cpp if we build it manually.

* fix: use force_pip for Windows ROCm bitsandbytes prebuilt wheel install

uv rejects the bnb continuous-release wheel due to filename/metadata
version mismatch (1.33.7.preview vs 0.50.0.dev0). Switch to force_pip=True
(pip bypass) instead of the UV_SKIP_WHEEL_FILENAME_CHECK env var workaround
-- cleaner and consistent with how the Linux path handles it.

BNB_ROCM_VERSION is still set post-install to the detected DLL suffix so
the worker subprocess loads the correct libbitsandbytes_rocm{VER}.dll even
when torch.version.hip reports a newer HIP version than the wheel ships.

* fix: three small correctness fixes found in PR review

- _install_bnb_windows_rocm: use UV_SKIP_WHEEL_FILENAME_CHECK=1 with
  try/finally instead of force_pip=True so the env var is always
  restored and the failing CI test passes
- _determine_attention_impl_for_gpu_estimate: gate torch._C distributed
  stubs on IS_ROCM so Windows CUDA users keep the real extension
- install.ps1 amd-smi fallback: collect all gfx tokens and index by
  HIP_VISIBLE_DEVICES, matching the hipinfo path on multi-GPU hosts

* fix: stub torchao in export subprocess on Windows ROCm

On Windows, the ROCm build of PyTorch ships without the distributed
C extension (torch._C._distributed_c10d). torchao, which is pulled in
transitively by transformers.quantizers at import time, walks into
torch.distributed._functional_collectives -> distributed_c10d and
crashes with:

  No module named 'torch._C._distributed_c10d'; 'torch._C' is not a package

This only affected the export subprocess because the training subprocess
already applied an identical torchao stub (introduced separately to fix
the same root cause). The export subprocess had no such guard and died
during 'Importing Unsloth...' before any model loading could happen.

Fix: apply the same _StubSubpackageFinder / torchao stub pattern to the
export subprocess entry point, gated on Windows ROCm detection, before
any import of transformers or unsloth_zoo.

Root cause tracked in ROCm/TheRock#3284 (libuv / torch.distributed
missing on Windows ROCm builds).

Ref: https://github.com/ROCm/TheRock/issues/3284

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* install.sh, setup.sh: add GPU arch step logging to match PS1 scripts

Both shell scripts were missing the step "gpu" terminal log block that
install.ps1 and setup.ps1 emit. This adds equivalent output: GPU label
with gfx arch (e.g. "AMD ROCm (gfx1151)"), ROCm root path, hipconfig
version, and marketing name substep. Includes the same gfx arch detection
chain (rocminfo → amd-smi list → amd-smi static --asic), UNSLOTH_ROCM_GFX_ARCH
env override, and name-based arch inference table (Strix Halo/Point, RDNA 3/4)
as the PS1 versions. install.sh also replaces bare echo blocks for the AMD
ROCm and CPU-only cases with formatted substep output.

* Fix BNB_ROCM_VERSION gate, ROCm GPU mask preference, APU unified memory and Release build for PR #5301

- main.py: gate BNB_ROCM_VERSION on the rocm bnb DLL or HIP_PATH/ROCM_PATH instead of importing torch on every Windows host
- hardware.py: prefer HIP/ROCR visible-device masks only on ROCm hosts so a stale mask cannot override CUDA_VISIBLE_DEVICES on NVIDIA
- llama_cpp.py: set GGML_CUDA_ENABLE_UNIFIED_MEMORY=1 only for unified-memory APUs (gfx1150/gfx1151)
- setup.sh: pass -DCMAKE_BUILD_TYPE=Release for the HIP source build
- add test_amd_apu_unified_memory.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: guard recompile_limit + fix AMD VRAM monitor fallback

trainer.py: torch._dynamo.config.recompile_limit does not exist in
some ROCm torch builds (e.g. pytorch.org/whl/rocm6.2 wheels). Guard
the assignment so training doesn't crash on RDNA2/RDNA3.

hardware.py: when amd-smi/nvidia-smi is unavailable or returns no
usable data (HIP SDK-only Windows, Docker, unexpected JSON format),
the existing fallback used torch.cuda.memory_allocated() which is
process-specific and reads near-zero even with a fully loaded model.
Switch to torch.cuda.mem_get_info() via _torch_get_per_device_info()
which reports system-wide VRAM occupancy so the GPU monitor shows
real usage on all AMD systems without requiring amd-smi.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: Windows VRAM monitor via Performance Counter API

When amd-smi/nvidia-smi is unavailable on Windows, query dedicated GPU
VRAM via Windows Performance Counters (same source as Task Manager).
This gives system-wide cross-process usage, fixing the near-zero reading
caused by torch.cuda.mem_get_info only seeing the Studio server process.

Linux fallback path unchanged (mem_get_info is system-wide on ROCm).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: rename to _rocm_windows_perf_counter_vram_gb, scope to IS_ROCM

Function is AMD ROCm specific — amd-smi absent on Windows when only the
HIP SDK is installed. Scoped to IS_ROCM so NVIDIA Windows path is
untouched (nvidia-smi handles that case).

* fix: AMD VRAM monitor — Linux DRM sysfs + Windows perf counter

Linux: read /sys/class/drm/card*/device/mem_info_vram_used|total for
system-wide GPU memory across all processes. No tools required, always
present on Linux AMD systems.

Windows: Windows Performance Counter API (already added).

Both paths are gated on IS_ROCM and only fire when amd-smi is absent.
torch mem_get_info remains as last resort (process-local).

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: AMD GPU monitor — utilization, temperature, and power for Windows and Linux fallback paths

- Windows: GPU utilization via \GPU Engine(*engtype_3D*)\Utilization Percentage perf counter
- Windows: temperature and power via ADL (atiadlxx.dll, ships with Adrenalin)
- Linux: GPU utilization via DRM sysfs gpu_busy_percent
- Linux: temperature via hwmon temp1_input (millidegrees C)
- Linux: power via hwmon power1_average / power1_input (microwatts)

All paths are no-op fallbacks (None) when the source is unavailable.
Mirrors what nvidia-smi provides on the CUDA path.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix: remove ADL ctypes — does not support AMD iGPU (Strix Halo)

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: Erland366 <erland.pg366@gmail.com>
Co-authored-by: danielhanchen <michaelhan2050@gmail.com>
2026-05-29 22:29:56 -07:00

2593 lines
115 KiB
Python

"""Tests for AMD ROCm support across install pathways.
Verifies that ROCm detection and installation logic works correctly
WITHOUT breaking existing CUDA, CPU, macOS, and Windows pathways.
All tests use mocks -- no AMD hardware required.
"""
import importlib.util
import json
import os
import subprocess
import sys
from pathlib import Path
from unittest.mock import MagicMock, patch, PropertyMock
import pytest
# ── Load modules under test ──────────────────────────────────────────────────
PACKAGE_ROOT = Path(__file__).resolve().parents[3]
# install_llama_prebuilt.py
_PREBUILT_PATH = PACKAGE_ROOT / "studio" / "install_llama_prebuilt.py"
_PREBUILT_SPEC = importlib.util.spec_from_file_location(
"studio_install_llama_prebuilt", _PREBUILT_PATH
)
assert _PREBUILT_SPEC is not None and _PREBUILT_SPEC.loader is not None
prebuilt_mod = importlib.util.module_from_spec(_PREBUILT_SPEC)
sys.modules[_PREBUILT_SPEC.name] = prebuilt_mod
_PREBUILT_SPEC.loader.exec_module(prebuilt_mod)
HostInfo = prebuilt_mod.HostInfo
AssetChoice = prebuilt_mod.AssetChoice
PrebuiltFallback = prebuilt_mod.PrebuiltFallback
resolve_upstream_asset_choice = prebuilt_mod.resolve_upstream_asset_choice
runtime_patterns_for_choice = prebuilt_mod.runtime_patterns_for_choice
# install_python_stack.py
_STACK_PATH = PACKAGE_ROOT / "studio" / "install_python_stack.py"
_STACK_SPEC = importlib.util.spec_from_file_location(
"studio_install_python_stack", _STACK_PATH
)
assert _STACK_SPEC is not None and _STACK_SPEC.loader is not None
stack_mod = importlib.util.module_from_spec(_STACK_SPEC)
sys.modules[_STACK_SPEC.name] = stack_mod
_STACK_SPEC.loader.exec_module(stack_mod)
_detect_rocm_version = stack_mod._detect_rocm_version
_ensure_rocm_torch = stack_mod._ensure_rocm_torch
_has_rocm_gpu = stack_mod._has_rocm_gpu
_has_usable_nvidia_gpu = stack_mod._has_usable_nvidia_gpu
_ROCM_TORCH_INDEX = stack_mod._ROCM_TORCH_INDEX
_windows_rocm_index_url = stack_mod._windows_rocm_index_url
_detect_windows_gfx_arch = stack_mod._detect_windows_gfx_arch
_install_bnb_windows_rocm = stack_mod._install_bnb_windows_rocm
def _extract_sh_function_body(source: str, name: str) -> str:
"""Return the body of a shell function from `source` by brace matching.
Used by structural tests that need to assert ordering of helper
calls inside a specific function rather than across the whole
install.sh file.
"""
needle = f"{name}() {{"
start = source.find(needle)
if start < 0:
return ""
depth = 0
i = start + len(needle) - 1 # land on the opening brace
n = len(source)
while i < n:
ch = source[i]
if ch == "{":
depth += 1
elif ch == "}":
depth -= 1
if depth == 0:
return source[start : i + 1]
i += 1
return source[start:]
# ── Helper: build HostInfo for different scenarios ──────────────────────────
def nvidia_host(**overrides) -> HostInfo:
"""NVIDIA Linux x86_64 host."""
defaults = dict(
system = "Linux",
machine = "x86_64",
is_windows = False,
is_linux = True,
is_macos = False,
is_x86_64 = True,
is_arm64 = False,
nvidia_smi = "/usr/bin/nvidia-smi",
driver_cuda_version = (12, 6),
compute_caps = ["89"],
visible_cuda_devices = None,
has_physical_nvidia = True,
has_usable_nvidia = True,
has_rocm = False,
)
defaults.update(overrides)
return HostInfo(**defaults)
def rocm_host(**overrides) -> HostInfo:
"""AMD ROCm Linux x86_64 host (no NVIDIA)."""
defaults = dict(
system = "Linux",
machine = "x86_64",
is_windows = False,
is_linux = True,
is_macos = False,
is_x86_64 = True,
is_arm64 = False,
nvidia_smi = None,
driver_cuda_version = None,
compute_caps = [],
visible_cuda_devices = None,
has_physical_nvidia = False,
has_usable_nvidia = False,
has_rocm = True,
)
defaults.update(overrides)
return HostInfo(**defaults)
def cpu_host(**overrides) -> HostInfo:
"""CPU-only Linux x86_64 host."""
defaults = dict(
system = "Linux",
machine = "x86_64",
is_windows = False,
is_linux = True,
is_macos = False,
is_x86_64 = True,
is_arm64 = False,
nvidia_smi = None,
driver_cuda_version = None,
compute_caps = [],
visible_cuda_devices = None,
has_physical_nvidia = False,
has_usable_nvidia = False,
has_rocm = False,
)
defaults.update(overrides)
return HostInfo(**defaults)
def macos_host(**overrides) -> HostInfo:
"""macOS arm64 host."""
defaults = dict(
system = "Darwin",
machine = "arm64",
is_windows = False,
is_linux = False,
is_macos = True,
is_x86_64 = False,
is_arm64 = True,
nvidia_smi = None,
driver_cuda_version = None,
compute_caps = [],
visible_cuda_devices = None,
has_physical_nvidia = False,
has_usable_nvidia = False,
has_rocm = False,
)
defaults.update(overrides)
return HostInfo(**defaults)
def windows_host(**overrides) -> HostInfo:
"""Windows x86_64 host."""
defaults = dict(
system = "Windows",
machine = "amd64",
is_windows = True,
is_linux = False,
is_macos = False,
is_x86_64 = True,
is_arm64 = False,
nvidia_smi = None,
driver_cuda_version = None,
compute_caps = [],
visible_cuda_devices = None,
has_physical_nvidia = False,
has_usable_nvidia = False,
has_rocm = False,
)
defaults.update(overrides)
return HostInfo(**defaults)
def windows_rocm_host(**overrides) -> HostInfo:
"""Windows x86_64 host with ROCm."""
defaults = dict(
system = "Windows",
machine = "amd64",
is_windows = True,
is_linux = False,
is_macos = False,
is_x86_64 = True,
is_arm64 = False,
nvidia_smi = None,
driver_cuda_version = None,
compute_caps = [],
visible_cuda_devices = None,
has_physical_nvidia = False,
has_usable_nvidia = False,
has_rocm = True,
)
defaults.update(overrides)
return HostInfo(**defaults)
# ── Upstream asset fixture ───────────────────────────────────────────────────
LLAMA_TAG = "b8508"
UPSTREAM_ASSETS = {
f"llama-{LLAMA_TAG}-bin-ubuntu-x64.tar.gz": f"https://example.com/{LLAMA_TAG}-linux-cpu.tar.gz",
f"llama-{LLAMA_TAG}-bin-ubuntu-rocm-7.2-x64.tar.gz": f"https://example.com/{LLAMA_TAG}-linux-rocm.tar.gz",
f"llama-{LLAMA_TAG}-bin-win-cpu-x64.zip": f"https://example.com/{LLAMA_TAG}-win-cpu.zip",
f"llama-{LLAMA_TAG}-bin-win-cuda-12.4-x64.zip": f"https://example.com/{LLAMA_TAG}-win-cuda.zip",
f"llama-{LLAMA_TAG}-bin-win-hip-radeon-x64.zip": f"https://example.com/{LLAMA_TAG}-win-hip.zip",
f"llama-{LLAMA_TAG}-bin-macos-arm64.tar.gz": f"https://example.com/{LLAMA_TAG}-macos-arm64.tar.gz",
f"llama-{LLAMA_TAG}-bin-macos-x64.tar.gz": f"https://example.com/{LLAMA_TAG}-macos-x64.tar.gz",
}
# =============================================================================
# TEST: install_llama_prebuilt.py -- resolve_upstream_asset_choice
# =============================================================================
class TestResolveUpstreamAssetChoice:
"""Verify that the asset selection logic picks the right binary for each platform."""
@patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
def test_nvidia_linux_gets_cpu_asset(self, mock_assets):
"""NVIDIA host should NOT hit the ROCm path -- gets CPU asset (CUDA handled elsewhere)."""
host = nvidia_host()
choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
assert choice.install_kind == "linux-cpu"
assert "ubuntu-x64" in choice.name
assert "rocm" not in choice.name
@patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
def test_rocm_linux_gets_rocm_prebuilt(self, mock_assets):
"""AMD ROCm Linux host should get the ROCm prebuilt."""
host = rocm_host()
choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
assert choice.install_kind == "linux-rocm"
assert "rocm" in choice.name
@patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
def test_cpu_linux_gets_cpu_asset(self, mock_assets):
"""CPU-only Linux host should get CPU asset."""
host = cpu_host()
choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
assert choice.install_kind == "linux-cpu"
assert "ubuntu-x64" in choice.name
@patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
def test_macos_arm64_gets_macos_asset(self, mock_assets):
"""macOS arm64 host should get macOS asset."""
host = macos_host()
choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
assert choice.install_kind == "macos-arm64"
assert "macos-arm64" in choice.name
@patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
def test_windows_cpu_gets_cpu_asset(self, mock_assets):
"""Windows CPU-only host should get Windows CPU asset."""
host = windows_host()
choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
assert choice.install_kind == "windows-cpu"
assert "win-cpu" in choice.name
@patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
def test_windows_rocm_gets_hip_asset(self, mock_assets):
"""Windows ROCm host should get Windows HIP asset."""
host = windows_rocm_host()
choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
assert choice.install_kind == "windows-hip"
assert "hip" in choice.name
@patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
def test_mixed_nvidia_rocm_prefers_nvidia(self, mock_assets):
"""Host with both NVIDIA and ROCm should use NVIDIA (CPU path here, CUDA elsewhere)."""
host = nvidia_host(has_rocm = True)
choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
# NVIDIA hosts go through the normal path (CUDA handled by resolve_linux_cuda_choice)
assert choice.install_kind == "linux-cpu"
assert "rocm" not in choice.name
@patch.object(prebuilt_mod, "github_release_assets")
def test_rocm_linux_no_prebuilt_falls_back(self, mock_assets):
"""AMD ROCm host should fall back to source build when no ROCm prebuilt exists."""
# Remove the ROCm asset from available assets
assets_without_rocm = {
k: v for k, v in UPSTREAM_ASSETS.items() if "rocm" not in k
}
mock_assets.return_value = assets_without_rocm
host = rocm_host()
with pytest.raises(PrebuiltFallback, match = "ROCm detected"):
resolve_upstream_asset_choice(host, LLAMA_TAG)
@patch.object(prebuilt_mod, "github_release_assets")
def test_windows_rocm_no_hip_falls_to_cpu(self, mock_assets):
"""Windows+ROCm with HIP prebuilt missing should fall through to CPU."""
assets_no_hip = {k: v for k, v in UPSTREAM_ASSETS.items() if "hip" not in k}
mock_assets.return_value = assets_no_hip
host = windows_rocm_host()
choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
assert choice.install_kind == "windows-cpu"
@patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
def test_macos_rocm_impossible_has_rocm_false(self, mock_assets):
"""macOS host should never have has_rocm=True in practice; verify it gets macOS asset."""
host = macos_host(has_rocm = True)
choice = resolve_upstream_asset_choice(host, LLAMA_TAG)
assert choice.install_kind == "macos-arm64"
@patch.object(prebuilt_mod, "github_release_assets", return_value = UPSTREAM_ASSETS)
def test_linux_aarch64_rocm_gets_prebuilt_fallback(self, mock_assets):
"""Linux aarch64 with ROCm -- no x86_64 match, should raise PrebuiltFallback."""
host = rocm_host(machine = "aarch64", is_x86_64 = False, is_arm64 = True)
with pytest.raises(PrebuiltFallback):
resolve_upstream_asset_choice(host, LLAMA_TAG)
# =============================================================================
# TEST: install_llama_prebuilt.py -- runtime_patterns_for_choice
# =============================================================================
class TestRuntimePatterns:
"""Verify runtime file patterns for all install kinds."""
def test_linux_cpu_patterns(self):
choice = AssetChoice(
repo = "", tag = "", name = "", url = "", source_label = "", install_kind = "linux-cpu"
)
patterns = runtime_patterns_for_choice(choice)
assert "llama-server" in patterns
assert "llama-quantize" in patterns
# Broad lib*.so* covers libllama, libggml, libmtmd, libggml-cpu-*,
# plus the libllama-<binary>-impl.so split that ggml-org/llama.cpp
# #23462 introduced between b9279 and b9283.
assert "lib*.so*" in patterns
def test_linux_cuda_patterns(self):
choice = AssetChoice(
repo = "", tag = "", name = "", url = "", source_label = "", install_kind = "linux-cuda"
)
patterns = runtime_patterns_for_choice(choice)
# libggml-cuda.so is matched by lib*.so* now.
assert "lib*.so*" in patterns
def test_linux_rocm_patterns(self):
choice = AssetChoice(
repo = "", tag = "", name = "", url = "", source_label = "", install_kind = "linux-rocm"
)
patterns = runtime_patterns_for_choice(choice)
# libggml-hip.so is matched by lib*.so* now.
assert "lib*.so*" in patterns
assert "llama-server" in patterns
def test_windows_hip_patterns(self):
choice = AssetChoice(
repo = "",
tag = "",
name = "",
url = "",
source_label = "",
install_kind = "windows-hip",
)
patterns = runtime_patterns_for_choice(choice)
# Narrowed from "*.exe" to the two binaries Studio actually
# invokes, mirroring the Linux/macOS pattern style.
assert "llama-server.exe" in patterns
assert "llama-quantize.exe" in patterns
assert "*.dll" in patterns
def test_macos_patterns(self):
choice = AssetChoice(
repo = "",
tag = "",
name = "",
url = "",
source_label = "",
install_kind = "macos-arm64",
)
patterns = runtime_patterns_for_choice(choice)
assert "lib*.dylib" in patterns
# =============================================================================
# TEST: install_llama_prebuilt.py -- HostInfo.has_rocm field
# =============================================================================
class TestHostInfoRocm:
"""Verify has_rocm field does not affect other HostInfo behavior."""
def test_has_rocm_default_false(self):
host = HostInfo(
system = "Linux",
machine = "x86_64",
is_windows = False,
is_linux = True,
is_macos = False,
is_x86_64 = True,
is_arm64 = False,
nvidia_smi = None,
driver_cuda_version = None,
compute_caps = [],
visible_cuda_devices = None,
has_physical_nvidia = False,
has_usable_nvidia = False,
)
assert host.has_rocm is False
def test_has_rocm_explicit_true(self):
host = rocm_host()
assert host.has_rocm is True
def test_nvidia_host_no_rocm(self):
host = nvidia_host()
assert host.has_rocm is False
assert host.has_usable_nvidia is True
def test_detect_host_has_rocm_detection_logic(self):
"""detect_host() should have ROCm GPU detection logic."""
import inspect
source = inspect.getsource(prebuilt_mod.detect_host)
# Must probe for actual GPU, not just tool presence
assert "rocminfo" in source or "amd-smi" in source
def test_detect_host_windows_rocm_detection(self):
"""detect_host() source should have Windows-specific ROCm GPU detection."""
import inspect
source = inspect.getsource(prebuilt_mod.detect_host)
assert "hipinfo" in source or "amd-smi" in source
# =============================================================================
# TEST: install_python_stack.py -- _detect_rocm_version
# =============================================================================
class TestDetectRocmVersion:
"""Verify ROCm version detection from various sources."""
def test_no_rocm_returns_none(self, tmp_path):
"""No ROCm installed should return None."""
with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path / "nonexistent")}):
with patch("shutil.which", return_value = None):
result = _detect_rocm_version()
assert result is None
def test_version_from_file(self, tmp_path):
"""Reads version from /opt/rocm/.info/version."""
info_dir = tmp_path / ".info"
info_dir.mkdir()
(info_dir / "version").write_text("7.1.0-12345\n")
with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
result = _detect_rocm_version()
assert result == (7, 1)
def test_version_62(self, tmp_path):
"""Reads ROCm 6.2 version."""
info_dir = tmp_path / ".info"
info_dir.mkdir()
(info_dir / "version").write_text("6.2.0\n")
with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
result = _detect_rocm_version()
assert result == (6, 2)
def test_hipconfig_fallback(self, tmp_path):
"""Falls back to hipconfig --version when file not found."""
with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path / "nonexistent")}):
mock_result = MagicMock()
mock_result.returncode = 0
mock_result.stdout = b"6.3.21234.2\n"
with patch("shutil.which", return_value = "/usr/bin/hipconfig"):
with patch("subprocess.run", return_value = mock_result):
result = _detect_rocm_version()
assert result == (6, 3)
def test_empty_version_file(self, tmp_path):
"""Empty version file should return None."""
info_dir = tmp_path / ".info"
info_dir.mkdir()
(info_dir / "version").write_text("")
with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
with patch("shutil.which", return_value = None):
result = _detect_rocm_version()
assert result is None
def test_version_with_epoch_prefix(self, tmp_path):
"""Debian epoch prefix (2:6.2.0) -- version file has no epoch, so should parse."""
info_dir = tmp_path / ".info"
info_dir.mkdir()
# Version files don't typically have epoch prefix, but lib/rocm_version might
(info_dir / "version").write_text("6.2.0\n")
with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
result = _detect_rocm_version()
assert result == (6, 2)
def test_multiple_version_sources_first_wins(self, tmp_path):
"""When both .info/version and lib/rocm_version exist, first found wins."""
info_dir = tmp_path / ".info"
info_dir.mkdir()
(info_dir / "version").write_text("7.1.0\n")
lib_dir = tmp_path / "lib"
lib_dir.mkdir()
(lib_dir / "rocm_version").write_text("6.3.0\n")
with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path)}):
result = _detect_rocm_version()
assert result == (7, 1) # .info/version checked first
def test_hipconfig_multiline_output(self, tmp_path):
"""hipconfig with multi-line output -- should use first line."""
with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path / "nonexistent")}):
mock_result = MagicMock()
mock_result.returncode = 0
mock_result.stdout = b"6.3.21234.2\nSome extra info\n"
with patch("shutil.which", return_value = "/usr/bin/hipconfig"):
with patch("subprocess.run", return_value = mock_result):
result = _detect_rocm_version()
assert result == (6, 3)
def test_hipconfig_timeout(self, tmp_path):
"""hipconfig that times out should return None."""
with patch.dict(os.environ, {"ROCM_PATH": str(tmp_path / "nonexistent")}):
with patch("shutil.which", return_value = "/usr/bin/hipconfig"):
with patch(
"subprocess.run",
side_effect = subprocess.TimeoutExpired("hipconfig", 5),
):
result = _detect_rocm_version()
assert result is None
# =============================================================================
# TEST: install_python_stack.py -- _ensure_rocm_torch
# =============================================================================
class TestEnsureRocmTorch:
"""Verify ROCm torch reinstall logic."""
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
def test_no_rocm_skips(self, mock_nvidia, mock_pip):
"""No ROCm toolchain should skip entirely."""
with patch("os.path.isdir", return_value = False):
with patch("shutil.which", return_value = None):
_ensure_rocm_torch()
mock_pip.assert_not_called()
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
@patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
@patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 1))
def test_torch_already_has_cuda_skips(
self, mock_ver, mock_gpu, mock_nvidia, mock_pip
):
"""If torch already has CUDA, should skip ROCm reinstall."""
mock_probe = MagicMock()
mock_probe.returncode = 0
mock_probe.stdout = b"12.6\n" # CUDA version string
with patch("os.path.isdir", return_value = True):
with patch("subprocess.run", return_value = mock_probe):
_ensure_rocm_torch()
mock_pip.assert_not_called()
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
@patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
@patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 1))
def test_torch_already_has_hip_skips(
self, mock_ver, mock_gpu, mock_nvidia, mock_pip
):
"""If torch already has HIP, should skip ROCm reinstall."""
mock_probe = MagicMock()
mock_probe.returncode = 0
mock_probe.stdout = b"7.1.12345\n" # HIP version string
with patch("os.path.isdir", return_value = True):
with patch("subprocess.run", return_value = mock_probe):
_ensure_rocm_torch()
mock_pip.assert_not_called()
@patch.object(stack_mod, "IS_WINDOWS", False)
@patch.object(stack_mod, "pip_install_try", return_value = True)
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
@patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
@patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 1))
def test_cpu_torch_gets_rocm_reinstall(
self, mock_ver, mock_gpu, mock_nvidia, mock_pip, mock_pip_try
):
"""CPU-only torch on ROCm host should trigger reinstall."""
mock_probe = MagicMock()
mock_probe.returncode = 0
mock_probe.stdout = b"\n" # empty = no GPU backend
with patch("os.path.isdir", return_value = True):
with patch("subprocess.run", return_value = mock_probe):
_ensure_rocm_torch()
# Should install torch via pip_install and bitsandbytes via pip_install_try.
assert mock_pip.call_count == 1
assert "rocm7.1" in str(mock_pip.call_args_list[0])
assert mock_pip_try.call_count >= 1
assert "bitsandbytes" in str(mock_pip_try.call_args_list[0])
assert mock_pip_try.call_args.kwargs["force_pip"] is True
@patch.object(stack_mod, "IS_WINDOWS", False)
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
@patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
@patch.object(stack_mod, "_detect_rocm_version", return_value = (6, 3))
def test_rocm_63_selects_correct_tag(
self, mock_ver, mock_gpu, mock_nvidia, mock_pip
):
"""ROCm 6.3 should select rocm6.3 tag."""
mock_probe = MagicMock()
mock_probe.returncode = 0
mock_probe.stdout = b"\n"
with patch("os.path.isdir", return_value = True):
with patch("subprocess.run", return_value = mock_probe):
_ensure_rocm_torch()
torch_call = mock_pip.call_args_list[0]
assert "rocm6.3" in str(torch_call)
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
@patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
@patch.object(stack_mod, "_detect_rocm_version", return_value = (5, 0))
def test_old_rocm_skips(self, mock_ver, mock_gpu, mock_nvidia, mock_pip):
"""ROCm version too old (below 6.0) should skip."""
mock_probe = MagicMock()
mock_probe.returncode = 0
mock_probe.stdout = b"\n"
with patch("os.path.isdir", return_value = True):
with patch("subprocess.run", return_value = mock_probe):
_ensure_rocm_torch()
mock_pip.assert_not_called()
@patch.object(stack_mod, "IS_WINDOWS", False)
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
@patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
@patch.object(stack_mod, "_detect_rocm_version", return_value = None)
def test_version_unreadable_prints_warning(
self, mock_ver, mock_gpu, mock_nvidia, mock_pip, capsys
):
"""ROCm detected but version unreadable should print warning and skip."""
with patch("os.path.isdir", return_value = True):
_ensure_rocm_torch()
mock_pip.assert_not_called()
captured = capsys.readouterr()
assert "unreadable" in captured.out
@patch.object(stack_mod, "IS_WINDOWS", False)
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
@patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
@patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 2))
def test_rocm_72_selects_72_tag(self, mock_ver, mock_gpu, mock_nvidia, mock_pip):
"""ROCm 7.2 should select rocm7.2 tag (now in mapping with torch 2.11.0)."""
mock_probe = MagicMock()
mock_probe.returncode = 0
mock_probe.stdout = b"\n"
with patch("os.path.isdir", return_value = True):
with patch("subprocess.run", return_value = mock_probe):
_ensure_rocm_torch()
torch_call = mock_pip.call_args_list[0]
assert "rocm7.2" in str(torch_call)
@patch.object(stack_mod, "IS_WINDOWS", False)
@patch.object(stack_mod, "pip_install_try", return_value = True)
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
@patch.object(stack_mod, "_has_rocm_gpu", return_value = True)
@patch.object(stack_mod, "_detect_rocm_version", return_value = (7, 1))
def test_probe_timeout_triggers_reinstall(
self, mock_ver, mock_gpu, mock_nvidia, mock_pip, mock_pip_try
):
"""Probe subprocess timeout should not crash; should proceed to reinstall."""
with patch("os.path.isdir", return_value = True):
with patch(
"subprocess.run", side_effect = subprocess.TimeoutExpired("python", 30)
):
_ensure_rocm_torch()
# If probe times out, the function should treat torch as unusable and reinstall
# both torch (via pip_install) and bitsandbytes (via pip_install_try).
assert mock_pip.call_count == 1
assert "rocm7.1" in str(mock_pip.call_args_list[0])
assert mock_pip_try.call_count >= 1
assert mock_pip_try.call_args.kwargs["force_pip"] is True
@patch.object(stack_mod, "pip_install")
@patch.object(stack_mod, "_has_usable_nvidia_gpu", return_value = False)
@patch.object(stack_mod, "_has_rocm_gpu", return_value = False)
def test_no_gpu_with_rocm_tools_skips(self, mock_gpu, mock_nvidia, mock_pip):
"""ROCm tools present but no actual AMD GPU should skip entirely."""
with patch("os.path.isdir", return_value = True):
_ensure_rocm_torch()
mock_pip.assert_not_called()
# =============================================================================
# TEST: install_python_stack.py -- _ROCM_TORCH_INDEX mapping
# =============================================================================
class TestRocmTorchIndex:
"""Verify the ROCm version -> torch index tag mapping."""
def test_mapping_is_sorted_descending(self):
"""Keys should be in descending order for the next() iteration to work."""
keys = list(_ROCM_TORCH_INDEX.keys())
assert keys == sorted(keys, reverse = True)
def test_rocm_72_in_mapping(self):
"""ROCm 7.2 should be in the active mapping (torch 2.11.0 now supported)."""
assert (7, 2) in _ROCM_TORCH_INDEX
assert _ROCM_TORCH_INDEX[(7, 2)] == "rocm7.2"
def test_rocm_71_maps_correctly(self):
assert _ROCM_TORCH_INDEX[(7, 1)] == "rocm7.1"
def test_rocm_63_maps_correctly(self):
assert _ROCM_TORCH_INDEX[(6, 3)] == "rocm6.3"
def test_rocm_60_maps_correctly(self):
assert _ROCM_TORCH_INDEX[(6, 0)] == "rocm6.0"
def test_all_tags_use_download_pytorch(self):
"""All tags should be for download.pytorch.org, not repo.radeon.com."""
for tag in _ROCM_TORCH_INDEX.values():
assert tag.startswith("rocm")
assert "radeon" not in tag
def test_newer_rocm_selects_best_match(self):
"""ROCm 7.2 (now in map) should select rocm7.2 directly."""
ver = (7, 2)
tag = next(
(
t
for (maj, mn), t in sorted(_ROCM_TORCH_INDEX.items(), reverse = True)
if ver >= (maj, mn)
),
None,
)
assert tag == "rocm7.2"
def test_rocm_64_selects_64(self):
ver = (6, 4)
tag = next(
(
t
for (maj, mn), t in sorted(_ROCM_TORCH_INDEX.items(), reverse = True)
if ver >= (maj, mn)
),
None,
)
assert tag == "rocm6.4"
# =============================================================================
# TEST: hardware.py -- IS_ROCM flag and detect_hardware
# =============================================================================
class TestHardwareRocmFlag:
"""Verify IS_ROCM flag behavior without importing the full hardware module."""
def test_hardware_py_has_is_rocm(self):
"""hardware.py should define IS_ROCM."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
assert "IS_ROCM: bool" in source and "False" in source
def test_hardware_py_sets_is_rocm_on_hip(self):
"""detect_hardware() should set IS_ROCM when torch.version.hip is set."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
assert 'torch.version, "hip"' in source or "torch.version.hip" in source
def test_hardware_py_still_returns_cuda_for_rocm(self):
"""DeviceType should remain CUDA even on ROCm -- no DeviceType.ROCM."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
# Ensure ROCM is NOT a DeviceType member
enum_section = source.split("class DeviceType")[1].split("\n\n")[0]
assert "ROCM" not in enum_section
def test_hardware_py_has_rocm_in_package_versions(self):
"""get_package_versions() should include 'rocm' key."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
assert '"rocm"' in source
def test_hardware_py_device_type_cuda_references_intact(self):
"""All existing DeviceType.CUDA references should still be present."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
# Key functions that must still reference DeviceType.CUDA
assert "DeviceType.CUDA" in source
assert "DEVICE = DeviceType.CUDA" in source
def test_is_rocm_exported_from_init(self):
"""IS_ROCM should be exported from hardware __init__.py."""
init_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "__init__.py"
)
source = init_path.read_text(encoding = "utf-8")
assert "IS_ROCM" in source
def test_is_rocm_in_all_list(self):
"""IS_ROCM should be in __all__ list in __init__.py."""
init_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "__init__.py"
)
source = init_path.read_text(encoding = "utf-8")
# Extract __all__ section
assert '"IS_ROCM"' in source
def test_get_package_versions_returns_rocm_key(self):
"""get_package_versions() source should return both 'cuda' and 'rocm' keys."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
# Find the get_package_versions function body
func_start = source.find("def get_package_versions")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert '"cuda"' in func_body
assert '"rocm"' in func_body
def test_distributed_stubs_cover_is_torchelastic_launched(self):
"""_determine_attention_impl_for_gpu_estimate must stub is_torchelastic_launched.
resolve_attention_implementation calls is_torchelastic_launched() on
Windows ROCm where torch.distributed ships without that helper, causing
a warning: 'module torch.distributed has no attribute is_torchelastic_launched'.
"""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
assert "is_torchelastic_launched" in source
def test_distributed_stubs_cover_core_helpers(self):
"""_determine_attention_impl_for_gpu_estimate must stub the four core distributed helpers."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
for attr in ("is_initialized", "is_available", "get_rank", "get_world_size"):
assert (
attr in source
), f"distributed stub for '{attr}' missing from hardware.py"
# =============================================================================
# TEST: tokenizer_utils.py -- error message
# =============================================================================
class TestTokenizerErrorMessage:
"""Verify the AMD error message is updated."""
def test_no_old_amd_message(self):
"""Old 'We do not support AMD' message should be gone."""
tu_path = PACKAGE_ROOT / "unsloth" / "tokenizer_utils.py"
source = tu_path.read_text(encoding = "utf-8")
assert "We do not support AMD" not in source
def test_new_message_has_docs_link(self):
"""New message should point to Unsloth AMD docs."""
tu_path = PACKAGE_ROOT / "unsloth" / "tokenizer_utils.py"
source = tu_path.read_text(encoding = "utf-8")
assert "docs.unsloth.ai" in source or "No GPU detected" in source
# =============================================================================
# TEST: install.sh -- structural checks
# =============================================================================
class TestInstallShStructure:
"""Verify install.sh structural properties without running it."""
def test_no_here_strings(self):
"""install.sh must not use <<< (not POSIX)."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
# <<< is bash-only; breaks dash
for i, line in enumerate(source.splitlines(), 1):
stripped = line.lstrip()
if stripped.startswith("#"):
continue
assert "<<<" not in line, f"install.sh:{i} uses non-POSIX <<< here-string"
def test_rocm_detection_present(self):
"""install.sh should have ROCm detection in get_torch_index_url."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
assert "amd-smi" in source
assert "rocm" in source.lower()
def test_cuda_precedence(self):
"""ROCm detection should only run when nvidia-smi is absent.
install.sh defines _has_amd_rocm_gpu and _has_usable_nvidia_gpu
helpers near each other (file-position order has no semantic
meaning), so check the runtime ordering inside
get_torch_index_url instead: NVIDIA branch runs first and the
AMD/ROCm branch only fires inside the `if [ -z "$_smi" ]`
block.
"""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
body = _extract_sh_function_body(source, "get_torch_index_url")
nvidia_call = body.find("_has_usable_nvidia_gpu")
no_nvidia_branch = body.find('if [ -z "$_smi" ]')
rocm_call = body.find("_has_amd_rocm_gpu")
assert (
nvidia_call >= 0
), "get_torch_index_url should call _has_usable_nvidia_gpu"
assert (
no_nvidia_branch >= 0
), "get_torch_index_url should gate ROCm on no-nvidia-smi"
assert (
rocm_call > no_nvidia_branch
), "ROCm detection should sit inside the 'no nvidia-smi' branch"
assert (
nvidia_call < no_nvidia_branch
), "NVIDIA detection should run before the no-nvidia-smi branch"
def test_bitsandbytes_amd_install(self):
"""install.sh should install bitsandbytes for AMD when ROCm detected."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
assert "bitsandbytes" in source
assert "rocm*)" in source # case pattern for ROCm URLs
def test_cpu_hint_mentions_amd(self):
"""CPU-only hint should mention AMD ROCm."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
assert "ROCm" in source
def test_rocm72_supported_future_capped(self):
"""ROCm 7.2 should pass through directly; 7.3+ falls back to rocm7.2."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
assert 'echo "$_base/rocm7.2"' in source # fallback for unknown future versions
# Allowlisted versions should pass through directly
assert "rocm6.*" in source
assert "rocm7.0" in source
assert "rocm7.1" in source
assert "rocm7.2" in source
def test_rocm_tag_validation_guard_exists(self):
"""install.sh should validate _rocm_tag with a case guard."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
assert "rocm[1-9]*.[0-9]*)" in source
assert '_rocm_tag=""' in source # rejection path
def test_dpkg_epoch_handling(self):
"""install.sh should strip Debian epoch prefix from dpkg-query output."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
assert "sed 's/^[0-9]*://' " in source or "sed 's/^[0-9]*://'" in source
def test_no_double_bracket_in_rocm_block(self):
"""ROCm detection block should not use [[ ]] (bash-only, not POSIX).
Note: [[:space:]], [[:digit:]] etc. are valid POSIX character classes, not bash [[ ]]."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
func_start = source.find("get_torch_index_url()")
func_end = source.find("\n}", func_start)
func_body = source[func_start:func_end]
import re
for i, line in enumerate(func_body.splitlines(), 1):
stripped = line.lstrip()
if stripped.startswith("#"):
continue
# Remove POSIX character classes [[:foo:]] before checking for [[ ]]
cleaned = re.sub(r"\[\[:[a-z]+:\]\]", "", line)
assert (
"[[" not in cleaned
), f"get_torch_index_url line {i} uses non-POSIX [["
def test_no_arithmetic_expansion_in_rocm_block(self):
"""ROCm detection block should not use (( )) (bash-only)."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
func_start = source.find("get_torch_index_url()")
func_end = source.find("\n}", func_start)
func_body = source[func_start:func_end]
for i, line in enumerate(func_body.splitlines(), 1):
stripped = line.lstrip()
if stripped.startswith("#"):
continue
assert (
"((" not in line or "))" not in line or "$(()" in line
), f"get_torch_index_url line {i} may use non-POSIX (( ))"
def test_macos_returns_cpu_before_rocm_check(self):
"""macOS should return CPU immediately (before any ROCm check)."""
sh_path = PACKAGE_ROOT / "install.sh"
source = sh_path.read_text(encoding = "utf-8")
func_start = source.find("get_torch_index_url()")
func_body = source[func_start:]
darwin_pos = func_body.find("Darwin")
rocm_pos = func_body.find("amd-smi")
assert darwin_pos < rocm_pos, "macOS check should come before ROCm detection"
# =============================================================================
# TEST: Live regression on current host (NVIDIA B200 expected)
# =============================================================================
class TestLiveRegression:
"""Live checks that run on the actual host -- skip if no NVIDIA GPU."""
def test_get_torch_index_url_returns_cuda_on_nvidia(self):
"""On an NVIDIA machine, get_torch_index_url should return a CUDA URL."""
import shutil
if not shutil.which("nvidia-smi"):
pytest.skip("No nvidia-smi available")
# Skip if nvidia-smi exists but does not actually list a GPU on this
# host (containers occasionally ship the binary without a driver).
check = subprocess.run(
[
"bash",
"-c",
"nvidia-smi -L 2>/dev/null | "
"awk '/^GPU[[:space:]]+[0-9]+:/{f=1} END{exit !f}'",
],
capture_output = True,
)
if check.returncode != 0:
pytest.skip("nvidia-smi is on PATH but no GPU is listed")
sh_path = PACKAGE_ROOT / "install.sh"
# get_torch_index_url calls _has_usable_nvidia_gpu and
# _has_amd_rocm_gpu, so all three function definitions must be
# in scope when we eval the extract.
extract_cmd = (
f"sed -n '/^_has_amd_rocm_gpu()/,/^}}$/p; "
f"/^_has_usable_nvidia_gpu()/,/^}}$/p; "
f"/^get_torch_index_url()/,/^}}$/p' '{sh_path}'"
)
result = subprocess.run(
["bash", "-c", f'eval "$({extract_cmd})"; get_torch_index_url'],
capture_output = True,
text = True,
timeout = 30,
)
if result.returncode != 0:
pytest.skip("Could not extract get_torch_index_url for live test")
url = result.stdout.strip()
assert "cu1" in url or "cuda" in url.lower(), f"Expected CUDA URL, got: {url}"
# =============================================================================
# TEST: worker.py -- ROCm Mamba/SSM source build path
# =============================================================================
# Load worker.py module
_WORKER_PATH = PACKAGE_ROOT / "studio" / "backend" / "core" / "training" / "worker.py"
# The wheel-probe subprocess was hoisted out of worker.py into wheel_utils
# during the wheel-resolver refactor; the probe script literal lives there.
_WHEEL_UTILS_PATH = PACKAGE_ROOT / "studio" / "backend" / "utils" / "wheel_utils.py"
class TestWorkerRocmMambaSsm:
"""Verify worker.py Mamba/SSM install logic on ROCm."""
def test_probe_returns_hip_version_field(self):
"""The wheel probe should include hip_version, and worker.py should
consume it."""
assert "hip_version" in _WHEEL_UTILS_PATH.read_text(encoding = "utf-8")
assert "hip_version" in _WORKER_PATH.read_text(encoding = "utf-8")
def test_probe_script_has_getattr_hip(self):
"""Probe script should use getattr for torch.version.hip (safe on CUDA)."""
source = _WHEEL_UTILS_PATH.read_text(encoding = "utf-8")
assert "getattr(torch.version, 'hip', None)" in source
def test_direct_wheel_url_returns_none_without_cuda_major(self):
"""_direct_wheel_url should return None when cuda_major is empty (ROCm)."""
# Load module for function access
_worker_spec = importlib.util.spec_from_file_location(
"test_worker", _WORKER_PATH
)
assert _worker_spec is not None and _worker_spec.loader is not None
worker_mod = importlib.util.module_from_spec(_worker_spec)
# Mock all the imports worker.py needs
sys.modules["structlog"] = MagicMock()
sys.modules["loggers"] = MagicMock()
sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
sys.modules["utils"] = MagicMock()
sys.modules["utils.hardware"] = MagicMock()
try:
_worker_spec.loader.exec_module(worker_mod)
except Exception:
pytest.skip("Could not load worker module in test environment")
env_rocm = {
"python_tag": "cp312",
"torch_mm": "2.6",
"cuda_major": "",
"hip_version": "7.1.12345",
"cxx11abi": "TRUE",
}
result = worker_mod._direct_wheel_url(
filename_prefix = "causal_conv1d",
package_version = "1.6.1",
release_tag = "v1.6.1.post4",
release_base_url = "https://github.com/Dao-AILab/causal-conv1d/releases/download",
env = env_rocm,
)
assert result is None
def test_hipcc_check_exists_in_source(self):
"""worker.py should check for hipcc before ROCm source builds."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "hipcc" in source
def test_rocm_source_build_status_message(self):
"""worker.py should send a specific status for ROCm source compilation."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "Compiling" in source and "from source for ROCm" in source
def test_rocm_build_failure_message(self):
"""worker.py should send a clear error on ROCm build failure."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "Failed to compile" in source and "for ROCm" in source
def test_timeout_on_install(self):
"""worker.py should have a timeout on pip install subprocess."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "TimeoutExpired" in source
assert "timeout" in source
# =============================================================================
# TEST: amd.py -- AMD GPU monitoring
# =============================================================================
class TestAmdGpuMonitoring:
"""Verify amd.py module structure and mock behavior."""
def test_amd_py_exists(self):
"""amd.py should exist in the hardware directory."""
amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
assert amd_path.exists()
def test_amd_py_has_required_functions(self):
"""amd.py should export the same function signatures as nvidia.py."""
amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
source = amd_path.read_text(encoding = "utf-8")
assert "def get_physical_gpu_count" in source
assert "def get_primary_gpu_utilization" in source
assert "def get_visible_gpu_utilization" in source
def test_amd_smi_json_parsing(self):
"""Verify _extract_gpu_metrics parses amd-smi JSON correctly."""
amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
_amd_spec = importlib.util.spec_from_file_location("test_amd", amd_path)
assert _amd_spec is not None and _amd_spec.loader is not None
amd_mod = importlib.util.module_from_spec(_amd_spec)
sys.modules["loggers"] = MagicMock()
sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
try:
_amd_spec.loader.exec_module(amd_mod)
except Exception:
pytest.skip("Could not load amd module in test environment")
# Simulate amd-smi metric JSON output
gpu_data = {
"usage": {"gfx_activity": "85"},
"temperature": {"edge": "72"},
"power": {
"current_socket_power": "200.5",
"power_cap": "300",
},
"vram": {
"vram_used": 8192, # MB
"vram_total": 16384, # MB
},
}
metrics = amd_mod._extract_gpu_metrics(gpu_data)
assert metrics["gpu_utilization_pct"] == 85.0
assert metrics["temperature_c"] == 72.0
assert metrics["power_draw_w"] == 200.5
assert metrics["power_limit_w"] == 300.0
assert metrics["vram_used_gb"] == round(8192 / 1024, 2)
assert metrics["vram_total_gb"] == round(16384 / 1024, 2)
assert metrics["vram_utilization_pct"] is not None
assert metrics["power_utilization_pct"] is not None
def test_amd_primary_gpu_with_mock(self, monkeypatch):
"""get_primary_gpu_utilization returns correct dict with mocked amd-smi."""
amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
_amd_spec = importlib.util.spec_from_file_location("test_amd2", amd_path)
assert _amd_spec is not None and _amd_spec.loader is not None
amd_mod = importlib.util.module_from_spec(_amd_spec)
sys.modules["loggers"] = MagicMock()
sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
try:
_amd_spec.loader.exec_module(amd_mod)
except Exception:
pytest.skip("Could not load amd module")
# _first_visible_amd_gpu_id() short-circuits to None when any of
# HIP / ROCR / CUDA_VISIBLE_DEVICES is set to "" or "-1". CI runners
# often unset CUDA at the env level by setting CUDA_VISIBLE_DEVICES
# to "" so the test must not inherit that.
for var in (
"HIP_VISIBLE_DEVICES",
"ROCR_VISIBLE_DEVICES",
"CUDA_VISIBLE_DEVICES",
):
monkeypatch.delenv(var, raising = False)
mock_json = json.dumps(
[
{
"usage": {"gfx_activity": "50"},
"temperature": {"edge": "65"},
"power": {"current_socket_power": "150", "power_cap": "250"},
"vram": {"vram_used": 4096, "vram_total": 16384},
}
]
)
mock_result = MagicMock()
mock_result.returncode = 0
mock_result.stdout = mock_json
with patch.object(subprocess, "run", return_value = mock_result):
result = amd_mod.get_primary_gpu_utilization()
assert result["available"] is True
assert result["gpu_utilization_pct"] == 50.0
assert result["temperature_c"] == 65.0
def test_amd_smi_not_found_returns_unavailable(self):
"""get_primary_gpu_utilization returns available=False when amd-smi is missing."""
amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
_amd_spec = importlib.util.spec_from_file_location("test_amd3", amd_path)
assert _amd_spec is not None and _amd_spec.loader is not None
amd_mod = importlib.util.module_from_spec(_amd_spec)
sys.modules["loggers"] = MagicMock()
sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
try:
_amd_spec.loader.exec_module(amd_mod)
except Exception:
pytest.skip("Could not load amd module")
with patch.object(subprocess, "run", side_effect = OSError("amd-smi not found")):
result = amd_mod.get_primary_gpu_utilization()
assert result["available"] is False
def test_amd_timeout_returns_unavailable(self):
"""get_primary_gpu_utilization handles timeout gracefully."""
amd_path = PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "amd.py"
_amd_spec = importlib.util.spec_from_file_location("test_amd4", amd_path)
assert _amd_spec is not None and _amd_spec.loader is not None
amd_mod = importlib.util.module_from_spec(_amd_spec)
sys.modules["loggers"] = MagicMock()
sys.modules["loggers"].get_logger = MagicMock(return_value = MagicMock())
try:
_amd_spec.loader.exec_module(amd_mod)
except Exception:
pytest.skip("Could not load amd module")
with patch.object(
subprocess,
"run",
side_effect = subprocess.TimeoutExpired("amd-smi", 5),
):
result = amd_mod.get_primary_gpu_utilization()
assert result["available"] is False
# =============================================================================
# TEST: hardware.py -- IS_ROCM branching to amd.py
# =============================================================================
class TestHardwareAmdBranching:
"""Verify hardware.py branches to amd.py when IS_ROCM is True."""
def test_hardware_imports_amd_module(self):
"""hardware.py should import from amd module when IS_ROCM."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
assert "from . import amd" in source
def test_hardware_branches_on_is_rocm_for_utilization(self):
"""get_gpu_utilization should dispatch to amd.py via _smi_query
when IS_ROCM, and the dispatcher itself must check IS_ROCM and
import the amd backend."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
func_start = source.find("def get_gpu_utilization")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert '_smi_query("get_primary_gpu_utilization"' in func_body
smi = source[
source.find("def _smi_query") : source.find(
"\ndef ", source.find("def _smi_query") + 1
)
]
assert "IS_ROCM" in smi
assert "from . import amd" in smi
def test_hardware_branches_on_is_rocm_for_visible(self):
"""get_visible_gpu_utilization should dispatch to amd.py via
_smi_query when IS_ROCM."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
func_start = source.find("def get_visible_gpu_utilization")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
# The dispatcher call may wrap onto multiple lines; allow whitespace
# between the open paren and the literal func name argument.
import re as _re
assert _re.search(r'_smi_query\(\s*"get_visible_gpu_utilization"', func_body)
smi = source[
source.find("def _smi_query") : source.find(
"\ndef ", source.find("def _smi_query") + 1
)
]
assert "IS_ROCM" in smi
assert "from . import amd" in smi
def test_hardware_branches_on_is_rocm_for_physical_count(self):
"""get_physical_gpu_count should try amd.py when IS_ROCM."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
func_start = source.find("def get_physical_gpu_count")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert "IS_ROCM" in func_body
assert "from . import amd" in func_body
# =============================================================================
# TEST: hardware.py -- apply_gpu_ids ROCm fallback (issue #5180)
# =============================================================================
class TestApplyGpuIdsRocmFallback:
"""Verify apply_gpu_ids sets HIP_VISIBLE_DEVICES on ROCm hosts even when
IS_ROCM is still False (worker subprocess before detect_hardware runs)."""
def test_apply_gpu_ids_falls_back_to_torch_version_hip(self):
"""apply_gpu_ids should probe torch.version.hip when IS_ROCM is False and no ROCm env vars are set."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
func_start = source.find("def apply_gpu_ids")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert 'getattr(_torch.version, "hip", None)' in func_body
def test_apply_gpu_ids_sets_hip_and_rocr_visible_devices(self):
"""apply_gpu_ids should set both HIP_VISIBLE_DEVICES and ROCR_VISIBLE_DEVICES on ROCm."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
func_start = source.find("def apply_gpu_ids")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert 'os.environ["HIP_VISIBLE_DEVICES"] = value' in func_body
assert 'os.environ["ROCR_VISIBLE_DEVICES"] = value' in func_body
def test_apply_gpu_ids_rocm_fallback_is_guarded_by_try_except(self):
"""torch import in apply_gpu_ids must be wrapped in try/except so a missing torch never crashes."""
hw_path = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
source = hw_path.read_text(encoding = "utf-8")
func_start = source.find("def apply_gpu_ids")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert "import torch as _torch" in func_body
assert "except Exception" in func_body
# =============================================================================
# TEST: install_python_stack.py -- Windows AMD warning
# =============================================================================
class TestWindowsRocmWarning:
"""Verify Windows AMD GPU detection and warning message."""
def test_windows_amd_warning_in_source(self):
"""install_python_stack.py should warn Windows AMD users."""
source = _STACK_PATH.read_text(encoding = "utf-8")
assert "AMD GPU detected" in source
def test_windows_amd_warning_checks_hipinfo_or_amdsmi(self):
"""Warning should check for hipinfo or amd-smi."""
source = _STACK_PATH.read_text(encoding = "utf-8")
assert "hipinfo" in source
assert "amd-smi" in source
def test_windows_amd_warning_has_docs_link(self):
"""Warning should include AMD docs link."""
source = _STACK_PATH.read_text(encoding = "utf-8")
assert "docs.unsloth.ai/get-started/install-and-update/amd" in source
# =============================================================================
# TEST: unsloth/kernels/utils.py -- is_rdna() expansion
# =============================================================================
class TestIsRdnaExpansion:
"""Verify is_rdna() covers RDNA2, RDNA3, RDNA3.5, RDNA4 architectures."""
def test_is_rdna_source_has_rdna2(self):
"""is_rdna() should include RDNA2 architectures."""
utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
source = utils_path.read_text(encoding = "utf-8")
func_start = source.find("def is_rdna()")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert "gfx1030" in func_body
assert "gfx1031" in func_body
assert "gfx1032" in func_body
assert "gfx1033" in func_body
assert "gfx1034" in func_body
assert "gfx1035" in func_body
assert "gfx1036" in func_body
def test_is_rdna_source_has_rdna3(self):
"""is_rdna() should include RDNA3 architectures."""
utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
source = utils_path.read_text(encoding = "utf-8")
func_start = source.find("def is_rdna()")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert "gfx1100" in func_body
assert "gfx1101" in func_body
assert "gfx1102" in func_body
assert "gfx1103" in func_body
def test_is_rdna_source_has_rdna35(self):
"""is_rdna() should include RDNA3.5 architectures."""
utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
source = utils_path.read_text(encoding = "utf-8")
func_start = source.find("def is_rdna()")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert "gfx1150" in func_body
assert "gfx1151" in func_body
assert "gfx1152" in func_body
def test_is_rdna_source_has_rdna4(self):
"""is_rdna() should include RDNA4 architectures."""
utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
source = utils_path.read_text(encoding = "utf-8")
func_start = source.find("def is_rdna()")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert "gfx1200" in func_body
assert "gfx1201" in func_body
def test_is_cdna_not_changed(self):
"""is_cdna() should remain unchanged (no RDNA architectures added)."""
utils_path = PACKAGE_ROOT / "unsloth" / "kernels" / "utils.py"
source = utils_path.read_text(encoding = "utf-8")
func_start = source.find("def is_cdna()")
func_body = source[func_start : source.find("\ndef ", func_start + 1)]
assert "gfx940" in func_body
assert "gfx941" in func_body
assert "gfx942" in func_body
assert "gfx950" in func_body
# RDNA architectures should NOT be in is_cdna
assert "gfx1030" not in func_body
assert "gfx1100" not in func_body
# =============================================================================
# TEST: install_python_stack.py -- _windows_rocm_index_url arch mapping
# =============================================================================
class TestWindowsRocmIndexUrl:
"""Verify GPU arch → AMD pip index URL mapping."""
def test_gfx1200_maps_to_gfx120x_all(self):
url = stack_mod._windows_rocm_index_url("gfx1200")
assert url is not None
assert "gfx120X-all" in url
def test_gfx1201_maps_to_gfx120x_all(self):
url = stack_mod._windows_rocm_index_url("gfx1201")
assert url is not None
assert "gfx120X-all" in url
def test_gfx1151_maps_to_gfx1151(self):
url = stack_mod._windows_rocm_index_url("gfx1151")
assert url is not None
assert "gfx1151" in url
def test_gfx1150_maps_to_gfx1150(self):
url = stack_mod._windows_rocm_index_url("gfx1150")
assert url is not None
assert "gfx1150" in url
def test_gfx1100_maps_to_gfx110x_all(self):
url = stack_mod._windows_rocm_index_url("gfx1100")
assert url is not None
assert "gfx110X-all" in url
def test_unknown_arch_returns_none(self):
assert stack_mod._windows_rocm_index_url("gfx9999") is None
def test_none_arch_returns_none(self):
assert stack_mod._windows_rocm_index_url(None) is None
def test_url_ends_with_slash(self):
"""AMD pip index URLs must end with / for --index-url compatibility."""
url = stack_mod._windows_rocm_index_url("gfx1200")
assert url is not None
assert url.endswith("/")
def test_base_url_uses_repo_amd_com_by_default(self):
url = stack_mod._windows_rocm_index_url("gfx1200")
assert url is not None
assert "repo.amd.com" in url
def test_mirror_env_var_overrides_base(self, monkeypatch):
monkeypatch.setenv(
"UNSLOTH_ROCM_WINDOWS_MIRROR", "https://my-mirror.example.com/rocm/whl"
)
# Reload module-level constant by calling helper directly
url = stack_mod._windows_rocm_index_url("gfx1200")
# The env var is read at module load time for _ROCM_WINDOWS_INDEX_BASE,
# so just verify the helper itself doesn't error.
assert url is not None
# =============================================================================
# TEST: install_python_stack.py -- _detect_windows_gfx_arch
# =============================================================================
class TestDetectWindowsGfxArch:
"""Verify hipinfo parsing for GPU arch detection on Windows."""
def test_returns_none_when_hipinfo_not_on_path(self):
with patch("shutil.which", return_value = None):
result = stack_mod._detect_windows_gfx_arch()
assert result is None
def test_parses_gcnarchname_from_hipinfo_output(self):
mock_result = MagicMock()
mock_result.returncode = 0
mock_result.stdout = b"gcnArchName : gfx1200\nsome other line\n"
with patch("shutil.which", return_value = "/usr/bin/hipinfo"):
with patch("subprocess.run", return_value = mock_result):
result = stack_mod._detect_windows_gfx_arch()
assert result == "gfx1200"
def test_returns_none_on_nonzero_returncode(self):
mock_result = MagicMock()
mock_result.returncode = 1
mock_result.stdout = b"gcnArchName : gfx1200\n"
with patch("shutil.which", return_value = "/usr/bin/hipinfo"):
with patch("subprocess.run", return_value = mock_result):
result = stack_mod._detect_windows_gfx_arch()
assert result is None
def test_returns_none_when_no_gcnarchname_in_output(self):
mock_result = MagicMock()
mock_result.returncode = 0
mock_result.stdout = b"deviceName : Radeon RX 9060 XT\n"
with patch("shutil.which", return_value = "/usr/bin/hipinfo"):
with patch("subprocess.run", return_value = mock_result):
result = stack_mod._detect_windows_gfx_arch()
assert result is None
def test_returns_none_on_timeout(self):
with patch("shutil.which", return_value = "/usr/bin/hipinfo"):
with patch(
"subprocess.run",
side_effect = subprocess.TimeoutExpired("hipinfo", 10),
):
result = stack_mod._detect_windows_gfx_arch()
assert result is None
def test_strips_whitespace_from_arch(self):
mock_result = MagicMock()
mock_result.returncode = 0
mock_result.stdout = b" gcnArchName : gfx1201 \n"
with patch("shutil.which", return_value = "/usr/bin/hipinfo"):
with patch("subprocess.run", return_value = mock_result):
result = stack_mod._detect_windows_gfx_arch()
assert result == "gfx1201"
# =============================================================================
# TEST: install_python_stack.py -- _install_bnb_windows_rocm
# =============================================================================
class TestInstallBnbWindowsRocm:
"""Verify AMD Windows BNB wheel install helper."""
def test_calls_pip_install_try_with_win_amd64_url(self):
"""Should call pip_install_try with the win_amd64 wheel URL."""
with patch.object(stack_mod, "pip_install_try", return_value = True) as mock_pip:
stack_mod._install_bnb_windows_rocm()
assert mock_pip.call_count == 1
call_args = str(mock_pip.call_args_list[0])
assert "bitsandbytes" in call_args
assert "win_amd64" in call_args
def test_sets_uv_skip_env_var_during_install(self):
"""UV_SKIP_WHEEL_FILENAME_CHECK must be '1' when pip_install_try runs."""
observed = {}
def _capture(*args, **kwargs):
observed["val"] = os.environ.get("UV_SKIP_WHEEL_FILENAME_CHECK")
return True
with patch.object(stack_mod, "pip_install_try", side_effect = _capture):
stack_mod._install_bnb_windows_rocm()
assert observed.get("val") == "1"
def test_restores_uv_skip_env_var_after_install(self):
"""UV_SKIP_WHEEL_FILENAME_CHECK should be removed after install if it wasn't set before."""
with patch.dict(os.environ, {}, clear = False):
os.environ.pop("UV_SKIP_WHEEL_FILENAME_CHECK", None)
with patch.object(stack_mod, "pip_install_try", return_value = True):
stack_mod._install_bnb_windows_rocm()
assert "UV_SKIP_WHEEL_FILENAME_CHECK" not in os.environ
def test_restores_previous_uv_skip_value(self):
"""If UV_SKIP_WHEEL_FILENAME_CHECK was already set, restore it afterwards."""
with patch.dict(os.environ, {"UV_SKIP_WHEEL_FILENAME_CHECK": "0"}):
with patch.object(stack_mod, "pip_install_try", return_value = True):
stack_mod._install_bnb_windows_rocm()
assert os.environ.get("UV_SKIP_WHEEL_FILENAME_CHECK") == "0"
def test_restores_env_even_if_install_raises(self):
"""UV_SKIP_WHEEL_FILENAME_CHECK must be cleaned up even on pip failure."""
with patch.dict(os.environ, {}, clear = False):
os.environ.pop("UV_SKIP_WHEEL_FILENAME_CHECK", None)
with patch.object(
stack_mod, "pip_install_try", side_effect = RuntimeError("pip failed")
):
try:
stack_mod._install_bnb_windows_rocm()
except RuntimeError:
pass
assert "UV_SKIP_WHEEL_FILENAME_CHECK" not in os.environ
def test_no_op_when_win_amd64_url_missing(self):
"""Should be silent no-op if win_amd64 key absent from _BNB_ROCM_PRERELEASE_URLS."""
with patch.object(stack_mod, "_BNB_ROCM_PRERELEASE_URLS", {}):
with patch.object(stack_mod, "pip_install_try") as mock_pip:
stack_mod._install_bnb_windows_rocm()
mock_pip.assert_not_called()
def test_sets_bnb_rocm_version_from_detected_dll(self):
"""BNB_ROCM_VERSION is set from the DLL detected after install."""
with patch.dict(os.environ, {}, clear = False):
os.environ.pop("BNB_ROCM_VERSION", None)
with patch.object(stack_mod, "pip_install_try", return_value = True):
with patch.object(
stack_mod, "_detect_bnb_rocm_dll_ver", return_value = "72"
):
stack_mod._install_bnb_windows_rocm()
assert os.environ.get("BNB_ROCM_VERSION") == "72"
def test_sets_bnb_rocm_version_from_newer_dll(self):
"""If AMD ships a newer DLL (e.g. rocm713.dll), that version is used."""
with patch.dict(os.environ, {}, clear = False):
os.environ.pop("BNB_ROCM_VERSION", None)
with patch.object(stack_mod, "pip_install_try", return_value = True):
with patch.object(
stack_mod, "_detect_bnb_rocm_dll_ver", return_value = "713"
):
stack_mod._install_bnb_windows_rocm()
assert os.environ.get("BNB_ROCM_VERSION") == "713"
def test_falls_back_to_72_when_detection_fails(self):
"""Falls back to '72' when DLL detection returns None."""
with patch.dict(os.environ, {}, clear = False):
os.environ.pop("BNB_ROCM_VERSION", None)
with patch.object(stack_mod, "pip_install_try", return_value = True):
with patch.object(
stack_mod, "_detect_bnb_rocm_dll_ver", return_value = None
):
stack_mod._install_bnb_windows_rocm()
assert os.environ.get("BNB_ROCM_VERSION") == "72"
def test_does_not_override_existing_bnb_rocm_version(self):
"""An explicit BNB_ROCM_VERSION in the caller's env must not be clobbered."""
with patch.dict(os.environ, {"BNB_ROCM_VERSION": "60"}):
with patch.object(stack_mod, "pip_install_try", return_value = True):
stack_mod._install_bnb_windows_rocm()
assert os.environ.get("BNB_ROCM_VERSION") == "60"
class TestDetectBnbRocmDllVer:
"""Unit tests for _detect_bnb_rocm_dll_ver()."""
def test_returns_none_when_bnb_not_installed(self):
"""Returns None if bitsandbytes is not importable."""
import importlib.util
with patch.object(importlib.util, "find_spec", return_value = None):
assert stack_mod._detect_bnb_rocm_dll_ver() is None
def test_detects_rocm72_dll(self, tmp_path):
"""Returns '72' when libbitsandbytes_rocm72.dll is present."""
(tmp_path / "libbitsandbytes_rocm72.dll").write_text("")
mock_spec = MagicMock()
mock_spec.submodule_search_locations = [str(tmp_path)]
import importlib.util
with patch.object(importlib.util, "find_spec", return_value = mock_spec):
assert stack_mod._detect_bnb_rocm_dll_ver() == "72"
def test_detects_rocm713_dll(self, tmp_path):
"""Returns '713' when libbitsandbytes_rocm713.dll is present."""
(tmp_path / "libbitsandbytes_rocm713.dll").write_text("")
mock_spec = MagicMock()
mock_spec.submodule_search_locations = [str(tmp_path)]
import importlib.util
with patch.object(importlib.util, "find_spec", return_value = mock_spec):
assert stack_mod._detect_bnb_rocm_dll_ver() == "713"
def test_returns_none_when_only_cuda_dlls(self, tmp_path):
"""Returns None when only CUDA DLLs are present (no ROCm DLL)."""
(tmp_path / "libbitsandbytes_cuda121.dll").write_text("")
mock_spec = MagicMock()
mock_spec.submodule_search_locations = [str(tmp_path)]
import importlib.util
with patch.object(importlib.util, "find_spec", return_value = mock_spec):
assert stack_mod._detect_bnb_rocm_dll_ver() is None
def test_picks_highest_suffix_when_multiple_dlls(self, tmp_path):
"""Returns the highest numeric suffix when multiple ROCm DLL variants exist.
Filesystem glob order is not guaranteed, so the function must not stop
at the first match — it must always return the highest one.
"""
(tmp_path / "libbitsandbytes_rocm72.dll").write_text("")
(tmp_path / "libbitsandbytes_rocm713.dll").write_text("")
mock_spec = MagicMock()
mock_spec.submodule_search_locations = [str(tmp_path)]
import importlib.util
with patch.object(importlib.util, "find_spec", return_value = mock_spec):
assert stack_mod._detect_bnb_rocm_dll_ver() == "713"
# =============================================================================
# TEST: install_python_stack.py -- UNSLOTH_ROCM_TORCH_INSTALLED early-return path
# =============================================================================
class TestRocmTorchInstalledEnvVar:
"""Verify UNSLOTH_ROCM_TORCH_INSTALLED=1 skips main install but still installs BNB."""
@staticmethod
def _ok_torch_probe(*a, **kw):
# subprocess.run probe returns 0 when torch imports as ROCm
rv = MagicMock()
rv.returncode = 0
return rv
@patch.object(stack_mod, "_install_bnb_windows_rocm")
@patch.object(stack_mod, "pip_install")
def test_env_var_skips_main_pip_install(self, mock_pip, mock_bnb):
"""UNSLOTH_ROCM_TORCH_INSTALLED=1 should not trigger torch pip_install."""
with (
patch.dict(os.environ, {"UNSLOTH_ROCM_TORCH_INSTALLED": "1"}),
patch.object(stack_mod.subprocess, "run", side_effect = self._ok_torch_probe),
):
stack_mod._ensure_rocm_torch()
mock_pip.assert_not_called()
@patch.object(stack_mod, "_install_bnb_windows_rocm")
@patch.object(stack_mod, "pip_install")
def test_env_var_calls_bnb_install(self, mock_pip, mock_bnb):
"""UNSLOTH_ROCM_TORCH_INSTALLED=1 should still call _install_bnb_windows_rocm."""
with (
patch.dict(os.environ, {"UNSLOTH_ROCM_TORCH_INSTALLED": "1"}),
patch.object(stack_mod.subprocess, "run", side_effect = self._ok_torch_probe),
):
stack_mod._ensure_rocm_torch()
mock_bnb.assert_called_once()
@patch.object(stack_mod, "_install_bnb_windows_rocm")
@patch.object(stack_mod, "pip_install")
def test_env_var_sets_rocm_windows_flag(self, mock_pip, mock_bnb):
"""UNSLOTH_ROCM_TORCH_INSTALLED=1 should set _rocm_windows_torch_installed."""
stack_mod._rocm_windows_torch_installed = False
with (
patch.dict(os.environ, {"UNSLOTH_ROCM_TORCH_INSTALLED": "1"}),
patch.object(stack_mod.subprocess, "run", side_effect = self._ok_torch_probe),
):
stack_mod._ensure_rocm_torch()
assert stack_mod._rocm_windows_torch_installed is True
@patch.object(stack_mod, "_install_bnb_windows_rocm")
@patch.object(stack_mod, "pip_install")
def test_env_var_falls_through_when_torch_missing(self, mock_pip, mock_bnb):
"""If the venv was wiped between runs, the stale env-var must not suppress reinstall."""
stack_mod._rocm_windows_torch_installed = False
def _bad_probe(*a, **kw):
rv = MagicMock()
rv.returncode = 1
return rv
with (
patch.dict(os.environ, {"UNSLOTH_ROCM_TORCH_INSTALLED": "1"}),
patch.object(stack_mod.subprocess, "run", side_effect = _bad_probe),
patch.object(stack_mod, "IS_WINDOWS", False),
patch.object(stack_mod, "IS_MACOS", True),
):
stack_mod._ensure_rocm_torch()
# macOS branch is the next exit -- but the point is the early-return did NOT fire.
mock_bnb.assert_not_called()
# =============================================================================
# TEST: worker.py -- Windows ROCm patches (source-level checks)
# =============================================================================
class TestWorkerWindowsRocmPatches:
"""Verify worker.py contains the required Windows ROCm runtime patches."""
def test_grouped_mm_dispatch_patch_present(self):
"""worker.py must register a _grouped_mm CUDA dispatch override."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert '_gm_lib.impl("_grouped_mm"' in source
def test_grouped_mm_patch_targets_cuda_dispatch_key(self):
"""The dispatch override must target the CUDA key (not CompositeImplicitAutograd)."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert '"_grouped_mm", _grouped_mm_safe_impl, "CUDA"' in source
def test_grouped_mm_lib_kept_alive(self):
"""The Library object must be stored to prevent GC clearing the registration."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "_WINDOWS_ROCM_GROUPED_MM_LIB" in source
def test_grouped_mm_handles_offs_grouped_case(self):
"""_grouped_mm fallback must handle the grouped (offs!=None) variant."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "offs_list" in source
assert "offs.tolist()" in source
def test_torchao_stub_uses_stub_type_meta(self):
"""Torchao stub must use _StubTypeMeta so isinstance() returns False not TypeError."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "_StubTypeMeta" in source
def test_stub_type_meta_has_instancecheck(self):
"""_StubTypeMeta must define __instancecheck__ returning False."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "__instancecheck__" in source
def test_stub_subpackage_finder_registered(self):
"""_StubSubpackageFinder must be appended to sys.meta_path."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "sys.meta_path.append(_StubSubpackageFinder())" in source
def test_torchao_key_submodules_pre_stubbed(self):
"""Key torchao submodules (dtypes, quantization) must be pre-stubbed."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "torchao.dtypes" in source
assert "torchao.quantization" in source
def test_torchdynamo_disabled_on_windows_rocm(self):
"""worker.py should disable dynamo on Windows ROCm as belt-and-suspenders."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "TORCHDYNAMO_DISABLE" in source
def test_bnb_rocm_version_set_on_windows_rocm(self):
"""worker.py must set BNB_ROCM_VERSION in the Windows ROCm section.
BNB auto-detects HIP version from torch.version.hip, which can mismatch
the DLL suffix in the AMD prerelease wheel. The worker must detect the
actual DLL suffix and override BNB's auto-detection before ML imports.
"""
source = _WORKER_PATH.read_text(encoding = "utf-8")
# Env var must be set
assert "BNB_ROCM_VERSION" in source
# Detection helper must be used
assert "_detect_bnb_rocm_dll_ver" in source or "libbitsandbytes_rocm" in source
# "72" must appear as the safe fallback
assert '"72"' in source or "'72'" in source
def test_bnb_rocm_version_set_before_ml_imports(self):
"""BNB_ROCM_VERSION must appear in section 1f, before section 2 ML imports."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
idx_bnb = source.find("BNB_ROCM_VERSION")
# Use the specific section-2 marker that appears in the worker process
# entry-point function (not the trainer helper which has its own "# ── 2.").
idx_sec2 = source.find("# ── 2. Now import ML libraries")
assert idx_bnb != -1, "BNB_ROCM_VERSION not found in worker.py"
assert (
idx_sec2 != -1
), "'# ── 2. Now import ML libraries' marker not found in worker.py"
assert idx_bnb < idx_sec2, (
"BNB_ROCM_VERSION must be set before section 2 ML imports "
f"(found at {idx_bnb}, section 2 at {idx_sec2})"
)
def test_grouped_mm_patch_guarded_by_windows_and_hip_check(self):
"""_grouped_mm patch must only apply on Windows + HIP torch."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
# Must check sys.platform == "win32"
assert 'sys.platform == "win32"' in source
# Must gate on HIP version — code uses getattr chain: "version" and "hip"
assert '"version"' in source and '"hip"' in source
def test_hip_ver_at_least_helper_defined(self):
"""_hip_ver_at_least helper must be defined inside the Windows ROCm block."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "def _hip_ver_at_least(major: int, minor: int)" in source
def test_grouped_mm_patch_gated_on_hip_lt_713(self):
"""_grouped_mm patch must be skipped on HIP >= 7.13 (AMD fixed the bug in ROCm 7.13)."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
# The guard must call _hip_ver_at_least with exactly (7, 13)
assert "_hip_ver_at_least(7, 13)" in source
# The patch must be inside the `if not` branch (negated guard)
assert "if not _hip_ver_at_least(7, 13):" in source
def test_grouped_mm_hip_713_skip_message_present(self):
"""worker.py must log a message when skipping the patch on HIP >= 7.13."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
assert "HIP >= 7.13" in source
assert "7.13" in source
def test_grouped_mm_patch_else_branch_present(self):
"""An else branch must follow the _hip_ver_at_least gate (skip path for 7.13+)."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
# There must be an else: after the if not _hip_ver_at_least(7, 13): block
gate_idx = source.find("if not _hip_ver_at_least(7, 13):")
assert gate_idx != -1, "Version gate not found in worker.py"
# The else: branch must appear after the gate
else_idx = source.find("else:", gate_idx)
assert else_idx != -1, "else: branch after _hip_ver_at_least gate not found"
def test_hip_ver_at_least_handles_amd_version_format(self):
"""_hip_ver_at_least must split on '.' and compare only major.minor (handles '7.13.99004')."""
source = _WORKER_PATH.read_text(encoding = "utf-8")
# Must split the version string and take the first two parts
assert 'split(".")[:2]' in source or ".split('.')[:2]" in source
# =============================================================================
# TEST: install_python_stack.py -- _ROCM_TORCH_PKG_SPECS mapping
# =============================================================================
class TestRocmTorchPkgSpecs:
"""Verify per-tag torch version specs are correct."""
def test_rocm72_has_torch_211(self):
"""rocm7.2 should specify torch 2.11.x."""
specs = stack_mod._ROCM_TORCH_PKG_SPECS.get("rocm7.2")
assert specs is not None
torch_spec = specs[0]
assert "2.11" in torch_spec
def test_default_caps_below_211(self):
"""Default spec (rocm7.1 and earlier) should cap below 2.11."""
specs = stack_mod._ROCM_TORCH_PKG_SPECS.get("_default")
assert specs is not None
torch_spec = specs[0]
assert "<2.11" in torch_spec
def test_specs_have_torch_vision_audio(self):
"""Each entry should be a 3-tuple: torch, torchvision, torchaudio."""
for tag, specs in stack_mod._ROCM_TORCH_PKG_SPECS.items():
assert len(specs) == 3, f"{tag}: expected (torch, torchvision, torchaudio)"
assert "torch" in specs[0]
assert "torchvision" in specs[1]
assert "torchaudio" in specs[2]
def test_gfx_to_amd_index_covers_rdna4(self):
"""_GFX_TO_AMD_INDEX_ARCH must cover gfx1200 and gfx1201 (RDNA 4)."""
mapping = stack_mod._GFX_TO_AMD_INDEX_ARCH
assert mapping.get("gfx1200") == "gfx120X-all"
assert mapping.get("gfx1201") == "gfx120X-all"
def test_gfx_to_amd_index_covers_strix_halo(self):
"""_GFX_TO_AMD_INDEX_ARCH must cover gfx1151 and gfx1150 (RDNA 3.5)."""
mapping = stack_mod._GFX_TO_AMD_INDEX_ARCH
assert mapping.get("gfx1151") == "gfx1151"
assert mapping.get("gfx1150") == "gfx1150"
def test_gfx_to_amd_index_covers_rdna3(self):
"""_GFX_TO_AMD_INDEX_ARCH must cover gfx1100-gfx1103 (RDNA 3)."""
mapping = stack_mod._GFX_TO_AMD_INDEX_ARCH
for arch in ("gfx1100", "gfx1101", "gfx1102", "gfx1103"):
assert mapping.get(arch) == "gfx110X-all", f"{arch} missing from mapping"
# =============================================================================
# TEST: setup.ps1 / install.ps1 -- Strix Halo gfx arch detection
# =============================================================================
_SETUP_PS1_PATH = PACKAGE_ROOT / "studio" / "setup.ps1"
_INSTALL_PS1_PATH = PACKAGE_ROOT / "install.ps1"
class TestStrixHaloGfxArchDetection:
"""Verify that setup.ps1 and install.ps1 have robust gfx arch detection
for Strix Halo / iGPU users who only have the HIP runtime (no hipinfo)."""
def test_amd_smi_static_asic_attempted_in_setup(self):
"""setup.ps1 must try 'amd-smi static --asic' when list output lacks gfx arch."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "static --asic" in source
def test_amd_smi_static_asic_attempted_in_install(self):
"""install.ps1 must try 'amd-smi static --asic' when list output lacks gfx arch."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "static --asic" in source
def test_env_var_override_in_setup(self):
"""setup.ps1 must honour UNSLOTH_ROCM_GFX_ARCH as a manual arch override."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "UNSLOTH_ROCM_GFX_ARCH" in source
def test_env_var_override_in_install(self):
"""install.ps1 must honour UNSLOTH_ROCM_GFX_ARCH as a manual arch override."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "UNSLOTH_ROCM_GFX_ARCH" in source
def test_name_arch_table_covers_strix_halo_in_setup(self):
"""setup.ps1 name→arch table must map 890M / Strix Halo to gfx1151."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "gfx1151" in source
assert "890M" in source or "Strix Halo" in source
def test_name_arch_table_covers_strix_halo_in_install(self):
"""install.ps1 name→arch table must map 890M / Strix Halo to gfx1151."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "gfx1151" in source
assert "890M" in source or "Strix Halo" in source
def test_name_arch_table_covers_strix_point_in_setup(self):
"""setup.ps1 name→arch table must map 880M / Strix Point to gfx1150."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "gfx1150" in source
assert "880M" in source or "Strix Point" in source
def test_name_arch_table_covers_strix_point_in_install(self):
"""install.ps1 name→arch table must map 880M / Strix Point to gfx1150."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "gfx1150" in source
assert "880M" in source or "Strix Point" in source
def test_name_arch_table_covers_rdna3_phoenix_in_setup(self):
"""setup.ps1 name→arch table must map 780M / Phoenix to gfx1103."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "gfx1103" in source
assert "780M" in source or "Phoenix" in source
def test_wmi_does_not_set_hasrocm_in_setup(self):
"""WMI block in setup.ps1 must NOT set $HasROCm = $true (no runtime confirmation)."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
# Find the WMI block and confirm HasROCm is not set inside it
wmi_idx = source.find("Win32_VideoController")
assert wmi_idx != -1, "WMI block not found in setup.ps1"
# The nearest HasROCm = $true must not appear between the WMI block
# and the closing brace of that if-block. We check by confirming
# $HasROCm = $true does NOT appear within 300 chars of the WMI call.
wmi_context = source[wmi_idx : wmi_idx + 300]
assert "$HasROCm = $true" not in wmi_context
def test_gfx_arch_regex_parses_from_amd_smi_output(self):
"""Both files must use the gfx\\d+[a-z]? regex to parse arch from amd-smi output."""
for path in (_SETUP_PS1_PATH, _INSTALL_PS1_PATH):
source = path.read_text(encoding = "utf-8")
# The regex pattern used to match gfx arches
assert (
"gfx\\d+" in source or r"gfx\d+" in source
), f"gfx arch regex not found in {path.name}"
# =============================================================================
# TEST: HIP SDK tool path resolution via HIP_PATH / ROCM_PATH env vars
# =============================================================================
class TestHipSdkEnvPathResolution:
"""Verify that both install scripts resolve hipinfo/hipconfig via HIP_PATH
and ROCM_PATH when the tools are not on $PATH, and emit explicit warnings."""
# ── hipinfo resolution ────────────────────────────────────────────────────
def test_setup_checks_hip_path_for_hipinfo(self):
"""setup.ps1 must reference HIP_PATH when resolving hipinfo."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "HIP_PATH" in source
assert "hipinfo" in source
def test_install_checks_hip_path_for_hipinfo(self):
"""install.ps1 must reference HIP_PATH when resolving hipinfo."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "HIP_PATH" in source
assert "hipinfo" in source
def test_setup_checks_rocm_path_as_hipinfo_fallback(self):
"""setup.ps1 must also check ROCM_PATH as a secondary hipinfo fallback."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "ROCM_PATH" in source
# Confirm the fallback pattern: HIP_PATH ?? ROCM_PATH (or equivalent elseif)
assert "ROCM_PATH" in source and "HIP_PATH" in source
def test_install_checks_rocm_path_as_hipinfo_fallback(self):
"""install.ps1 must also check ROCM_PATH as a secondary hipinfo fallback."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "ROCM_PATH" in source
assert "ROCM_PATH" in source and "HIP_PATH" in source
def test_setup_resolves_hipinfo_via_bin_subdir(self):
"""setup.ps1 must join the env var root with 'bin\\hipinfo.exe'."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert r"bin\hipinfo.exe" in source
def test_install_resolves_hipinfo_via_bin_subdir(self):
"""install.ps1 must join the env var root with 'bin\\hipinfo.exe'."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert r"bin\hipinfo.exe" in source
# ── hipinfo not-on-PATH warning ───────────────────────────────────────────
def test_setup_warns_when_hipinfo_not_on_path(self):
"""setup.ps1 must warn when hipinfo is found via env var but not on PATH."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "hipinfo not on PATH" in source
def test_install_warns_when_hipinfo_not_on_path(self):
"""install.ps1 must warn when hipinfo is found via env var but not on PATH."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "hipinfo not on PATH" in source
# ── warn when HIP_PATH set but exe missing ────────────────────────────────
def test_setup_warns_when_hip_path_set_but_exe_missing(self):
"""setup.ps1 must warn when HIP_PATH is set but hipinfo.exe is not present."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
# The warning must mention that the SDK install may be incomplete
assert "incomplete" in source or "not found at" in source
def test_install_warns_when_hip_path_set_but_exe_missing(self):
"""install.ps1 must warn when HIP_PATH is set but hipinfo.exe is not present."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "incomplete" in source or "not found at" in source
# ── hipinfo runtime error warning ─────────────────────────────────────────
def test_setup_warns_on_hipinfo_nonzero_exit(self):
"""setup.ps1 must warn when hipinfo runs but returns a non-zero exit code."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "HIP runtime error" in source or "runtime error" in source.lower()
def test_install_warns_on_hipinfo_nonzero_exit(self):
"""install.ps1 must warn when hipinfo runs but returns a non-zero exit code."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "HIP runtime error" in source or "runtime error" in source.lower()
# ── hipconfig resolution ──────────────────────────────────────────────────
def test_setup_resolves_hipconfig_via_bin_subdir(self):
"""setup.ps1 must also fall back to HIP_PATH/bin/hipconfig.exe for version detection."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert r"bin\hipconfig.exe" in source
def test_install_resolves_hipconfig_via_bin_subdir(self):
"""install.ps1 must also fall back to HIP_PATH/bin/hipconfig.exe for version detection."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert r"bin\hipconfig.exe" in source
def test_setup_warns_when_hipconfig_not_on_path(self):
"""setup.ps1 must warn when hipconfig is found via env var but not on PATH."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "hipconfig not on PATH" in source
def test_install_warns_when_hipconfig_not_on_path(self):
"""install.ps1 must warn when hipconfig is found via env var but not on PATH."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "hipconfig not on PATH" in source
# ── PATH fix hint ─────────────────────────────────────────────────────────
def test_setup_provides_path_fix_hint(self):
"""setup.ps1 must tell the user how to add the HIP bin dir to PATH."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
# Should mention adding to PATH or SetEnvironmentVariable
assert "PATH" in source and (
"SetEnvironmentVariable" in source or "Add" in source
)
def test_install_provides_path_fix_hint(self):
"""install.ps1 must tell the user how to add the HIP bin dir to PATH."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "PATH" in source and (
"SetEnvironmentVariable" in source or "Add" in source
)
# =============================================================================
# TEST: HIP SDK detected substep -- path + hipconfig version shown in terminal
# =============================================================================
class TestHipSdkDetectedSubstep:
"""Verify that both scripts print HIP SDK path and full hipconfig version
as substeps under the gpu step when AMD ROCm is successfully detected."""
def test_setup_prints_hip_sdk_path_substep(self):
"""setup.ps1 must print an 'HIP SDK:' substep showing the resolved path."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "HIP SDK:" in source
def test_install_prints_hip_sdk_path_substep(self):
"""install.ps1 must print an 'HIP SDK:' substep showing the resolved path."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "HIP SDK:" in source
def test_setup_shows_hipconfig_full_version(self):
"""setup.ps1 must capture and display the full hipconfig version string."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "ROCmVersionFull" in source or "hipconfig:" in source
def test_install_shows_hipconfig_full_version(self):
"""install.ps1 must capture and display the full hipconfig version string."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "ROCmVersionFull" in source or "hipconfig:" in source
def test_setup_captures_full_version_not_just_major_minor(self):
"""setup.ps1 must store the raw hipconfig output line, not just major.minor."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "ROCmVersionFull" in source
def test_install_captures_full_version_not_just_major_minor(self):
"""install.ps1 must store the raw hipconfig output line, not just major.minor."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "ROCmVersionFull" in source
def test_setup_uses_hip_path_or_rocm_path_for_sdk_display(self):
"""setup.ps1 HIP SDK path substep must check HIP_PATH then ROCM_PATH."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "HIP_PATH" in source and "ROCM_PATH" in source
def test_install_uses_hip_path_or_rocm_path_for_sdk_display(self):
"""install.ps1 HIP SDK path substep must check HIP_PATH then ROCM_PATH."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "HIP_PATH" in source and "ROCM_PATH" in source
def test_setup_rocm_step_uses_full_version(self):
"""setup.ps1 'rocm' step label must prefer the full version string."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "ROCmVersionFull" in source and "rocm" in source
# =============================================================================
# TEST: install.sh -- Strix Halo rocm7.1 → rocm7.2 override
# =============================================================================
_INSTALL_SH_PATH = PACKAGE_ROOT / "install.sh"
_SETUP_SH_PATH = PACKAGE_ROOT / "studio" / "setup.sh"
class TestStrixRocm71Override:
"""Verify install.sh skips Radeon repo and routes to AMD arch-specific index
for gfx1151/gfx1150 when ROCm 7.1 would otherwise be selected (known _grouped_mm segfault).
AMD's repo.amd.com/rocm/whl/gfx1151/ serves torch 2.11+rocm7.13 which has the real fix."""
def test_strix_gfx_detection_in_install_sh(self):
"""install.sh must detect gfx1151 and gfx1150 for the override."""
source = _INSTALL_SH_PATH.read_text(encoding = "utf-8")
assert "gfx1151" in source and "gfx1150" in source
def test_rocm71_override_to_amd_arch_index_in_install_sh(self):
"""install.sh must override TORCH_INDEX_URL to AMD arch-specific index for Strix."""
source = _INSTALL_SH_PATH.read_text(encoding = "utf-8")
# The override must route to AMD's arch-specific index (repo.amd.com/rocm/whl)
assert "repo.amd.com/rocm/whl" in source
assert "_strix_gfx" in source
# The URL must incorporate the detected gfx arch so gfx1151 → .../gfx1151/
strix_idx = source.find("_amd_strix_base")
assert strix_idx != -1
ctx = source[strix_idx : strix_idx + 500]
assert "_strix_gfx" in ctx
def test_radeon_repo_bypassed_for_strix_in_install_sh(self):
"""install.sh must set _amd_gpu_radeon=false when Strix + ROCm 7.1 detected."""
source = _INSTALL_SH_PATH.read_text(encoding = "utf-8")
assert "_amd_gpu_radeon=false" in source
def test_strix_override_warns_with_moe_utils_reference(self):
"""install.sh must emit a [WARN] mentioning the moe_utils segfault."""
source = _INSTALL_SH_PATH.read_text(encoding = "utf-8")
assert "moe_utils" in source or "_grouped_mm" in source
def test_strix_override_only_fires_on_rocm71(self):
"""install.sh must scope the Strix override to rocm7.1 only (not rocm7.2+)."""
source = _INSTALL_SH_PATH.read_text(encoding = "utf-8")
strix_idx = source.find("_strix_gfx")
assert strix_idx != -1
# Look back for the rocm7.1 pattern within 600 chars before _strix_gfx
context_before = source[max(0, strix_idx - 2400) : strix_idx]
assert "rocm7.1" in context_before
def test_torch_constraint_updated_for_strix_amd_index(self):
"""install.sh must set TORCH_CONSTRAINT>=2.11 when routing Strix to AMD index."""
source = _INSTALL_SH_PATH.read_text(encoding = "utf-8")
assert "TORCH_CONSTRAINT" in source and "2.11" in source
def test_amd_rocm_mirror_env_var_respected(self):
"""install.sh must honour UNSLOTH_AMD_ROCM_MIRROR for air-gapped installs."""
source = _INSTALL_SH_PATH.read_text(encoding = "utf-8")
assert "UNSLOTH_AMD_ROCM_MIRROR" in source
def test_tauri_family_recognises_amd_arch_url(self):
"""_tauri_torch_index_family must return a rocm* family for AMD arch-specific URLs."""
source = _INSTALL_SH_PATH.read_text(encoding = "utf-8")
# The function must have a case branch for repo.amd.com/rocm/whl/gfx* URLs
assert "rocm/whl/gfx" in source
# =============================================================================
# TEST: setup.sh -- gcc-install-dir fix for Ubuntu 24.04 + ROCm 7.x clang-20
# =============================================================================
class TestSetupShGccInstallDir:
"""Verify setup.sh applies the --gcc-install-dir flag when building llama.cpp
with HIP on Ubuntu 24.04+ to work around ROCm 7.x clang-20 header path bug."""
def test_gcc_install_dir_search_loop_present(self):
"""setup.sh must iterate gcc versions 14→11 to find one with C++ headers."""
source = _SETUP_SH_PATH.read_text(encoding = "utf-8")
assert "_GCC_INSTALL_DIR" in source
assert "/usr/lib/gcc/x86_64-linux-gnu" in source
def test_gcc_install_dir_checks_include_dir(self):
"""setup.sh must check that the gcc dir has an 'include' subdirectory."""
source = _SETUP_SH_PATH.read_text(encoding = "utf-8")
assert "include" in source and "_GCC_INSTALL_DIR" in source
def test_gcc_install_dir_appended_to_cmake_hip_flags(self):
"""setup.sh must pass --gcc-install-dir via CMAKE_HIP_FLAGS."""
source = _SETUP_SH_PATH.read_text(encoding = "utf-8")
assert "CMAKE_HIP_FLAGS" in source
assert "gcc-install-dir" in source
def test_gcc_install_dir_only_applied_in_hip_build_block(self):
"""The --gcc-install-dir fix must only apply in the HIP/ROCm build branch."""
source = _SETUP_SH_PATH.read_text(encoding = "utf-8")
# GGML_HIP=ON must appear before gcc-install-dir in the source
hip_idx = source.find("GGML_HIP=ON")
gcc_idx = source.find("gcc-install-dir")
assert hip_idx != -1 and gcc_idx != -1
assert hip_idx < gcc_idx
def test_gcc_install_dir_logs_substep(self):
"""setup.sh must print a substep when the gcc install dir is resolved."""
source = _SETUP_SH_PATH.read_text(encoding = "utf-8")
assert "gcc install dir" in source or "GCC_INSTALL_DIR" in source
# =============================================================================
# TEST: main.py -- BNB_ROCM_VERSION server startup + distributed stubs
# =============================================================================
_MAIN_PY_PATH = PACKAGE_ROOT / "studio" / "backend" / "main.py"
_HARDWARE_PY_PATH = (
PACKAGE_ROOT / "studio" / "backend" / "utils" / "hardware" / "hardware.py"
)
class TestServerStartupRocmFixes:
"""Verify main.py sets BNB_ROCM_VERSION before any bitsandbytes import and
hardware.py injects torch._C._distributed_c10d stubs before torch.distributed."""
# ── BNB_ROCM_VERSION in server process ────────────────────────────────────
def test_main_py_sets_bnb_rocm_version(self):
"""main.py must set BNB_ROCM_VERSION in the server process before imports."""
source = _MAIN_PY_PATH.read_text(encoding = "utf-8")
assert "BNB_ROCM_VERSION" in source
def test_main_py_bnb_detection_scoped_to_win32(self):
"""main.py BNB_ROCM_VERSION logic must be inside the win32 platform guard."""
source = _MAIN_PY_PATH.read_text(encoding = "utf-8")
win32_idx = source.find('sys.platform == "win32"')
bnb_idx = source.find("BNB_ROCM_VERSION")
assert win32_idx != -1 and bnb_idx != -1
assert win32_idx < bnb_idx
def test_main_py_bnb_dll_detection_uses_glob(self):
"""main.py must scan for libbitsandbytes_rocm*.dll to find the right version."""
source = _MAIN_PY_PATH.read_text(encoding = "utf-8")
assert "libbitsandbytes_rocm" in source
def test_main_py_bnb_falls_back_to_72(self):
"""main.py must fall back to BNB_ROCM_VERSION='72' when no DLL is found."""
source = _MAIN_PY_PATH.read_text(encoding = "utf-8")
assert '"72"' in source or "'72'" in source
def test_main_py_bnb_only_set_when_not_already_in_env(self):
"""main.py must not override an existing BNB_ROCM_VERSION env var."""
source = _MAIN_PY_PATH.read_text(encoding = "utf-8")
assert '"BNB_ROCM_VERSION" not in os.environ' in source
# ── torch._C._distributed_c10d stubs in hardware.py ──────────────────────
def test_hardware_py_injects_distributed_c10d_stub(self):
"""hardware.py must inject torch._C._distributed_c10d into sys.modules."""
source = _HARDWARE_PY_PATH.read_text(encoding = "utf-8")
assert "_distributed_c10d" in source
def test_hardware_py_stub_injected_before_distributed_import(self):
"""The sys.modules stub must be injected BEFORE import torch.distributed."""
source = _HARDWARE_PY_PATH.read_text(encoding = "utf-8")
c10d_idx = source.find("_distributed_c10d")
dist_idx = source.find("import torch.distributed")
assert c10d_idx != -1 and dist_idx != -1
assert c10d_idx < dist_idx
def test_hardware_py_stub_uses_types_moduletype(self):
"""hardware.py must create the stub with types.ModuleType."""
source = _HARDWARE_PY_PATH.read_text(encoding = "utf-8")
assert "ModuleType" in source
def test_hardware_py_stub_scoped_to_win32(self):
"""hardware.py distributed stub injection must be gated on win32."""
source = _HARDWARE_PY_PATH.read_text(encoding = "utf-8")
assert 'platform == "win32"' in source or "win32" in source
def test_hardware_py_stub_exposes_fake_process_group(self):
"""hardware.py stub must set FakeProcessGroup so torch.distributed doesn't raise AttributeError."""
source = _HARDWARE_PY_PATH.read_text(encoding = "utf-8")
assert "FakeProcessGroup" in source
def test_hardware_py_stub_exposes_process_group(self):
"""hardware.py stub must set ProcessGroup on the c10d stub."""
source = _HARDWARE_PY_PATH.read_text(encoding = "utf-8")
assert "ProcessGroup" in source
def test_hardware_py_stub_uses_setattr_for_symbols(self):
"""hardware.py must use setattr to populate stub symbols dynamically."""
source = _HARDWARE_PY_PATH.read_text(encoding = "utf-8")
assert "setattr" in source
def test_hardware_py_stub_all_c10d_siblings_covered(self):
"""hardware.py must stub all three torch._C._distributed_* submodules."""
source = _HARDWARE_PY_PATH.read_text(encoding = "utf-8")
assert "_distributed_c10d" in source
assert "_distributed_autograd" in source
assert "_distributed_rpc" in source
# =============================================================================
# TEST: install.ps1 / setup.ps1 -- HipSdkInstalled flag (SDK found, device inaccessible)
# =============================================================================
class TestHipSdkInstalledButDeviceInaccessible:
"""Verify that when hipinfo is found but exits non-zero (device not ROCm-accessible),
both scripts distinguish this from 'HIP SDK not found' and emit the correct message."""
def test_install_ps1_has_hip_sdk_installed_flag(self):
"""install.ps1 must track HipSdkInstalled separately from HasROCm."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "HipSdkInstalled" in source
def test_setup_ps1_has_hip_sdk_installed_flag(self):
"""setup.ps1 must track HipSdkInstalled separately from HasROCm."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "HipSdkInstalled" in source
def test_install_ps1_sets_flag_when_hipinfo_binary_found(self):
"""install.ps1 must set HipSdkInstalled=true inside the 'if ($hipinfoExe)' block."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
# HipSdkInstalled must be assigned inside the hipinfoExe block
hipinfo_block_idx = source.find("if ($hipinfoExe)")
sdk_flag_idx = source.find("$HipSdkInstalled = $true", hipinfo_block_idx)
assert hipinfo_block_idx != -1 and sdk_flag_idx != -1
assert sdk_flag_idx > hipinfo_block_idx
def test_setup_ps1_sets_flag_when_hipinfo_binary_found(self):
"""setup.ps1 must set HipSdkInstalled=true inside the 'if ($hipinfoExe)' block."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
hipinfo_block_idx = source.find("if ($hipinfoExe)")
sdk_flag_idx = source.find("$HipSdkInstalled = $true", hipinfo_block_idx)
assert hipinfo_block_idx != -1 and sdk_flag_idx != -1
assert sdk_flag_idx > hipinfo_block_idx
def test_install_ps1_version_capture_runs_when_sdk_installed(self):
"""install.ps1 must capture hipconfig version when HipSdkInstalled even if HasROCm is false."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "HasROCm -or $HipSdkInstalled" in source or "$HipSdkInstalled" in source
def test_setup_ps1_version_capture_runs_when_sdk_installed(self):
"""setup.ps1 must capture hipconfig version when HipSdkInstalled even if HasROCm is false."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "HasROCm -or $HipSdkInstalled" in source or "$HipSdkInstalled" in source
def test_install_ps1_distinct_message_for_sdk_found_but_device_inaccessible(self):
"""install.ps1 must show 'not ROCm-accessible' message (not 'HIP SDK not found') when SDK present."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "not ROCm-accessible" in source
def test_setup_ps1_distinct_message_for_sdk_found_but_device_inaccessible(self):
"""setup.ps1 must show 'not ROCm-accessible' message (not 'HIP SDK not found') when SDK present."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "not ROCm-accessible" in source
def test_install_ps1_driver_guidance_in_sdk_found_branch(self):
"""install.ps1 must tell user this is a driver issue, not an SDK issue."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "driver issue" in source
def test_setup_ps1_driver_guidance_in_sdk_found_branch(self):
"""setup.ps1 must tell user this is a driver issue, not an SDK issue."""
source = _SETUP_PS1_PATH.read_text(encoding = "utf-8")
assert "driver issue" in source
def test_install_ps1_cpu_hint_distinguishes_driver_vs_no_sdk(self):
"""install.ps1 CPU-only hint must say 'GPU not ROCm-accessible' not 'require the HIP SDK' when SDK found."""
source = _INSTALL_PS1_PATH.read_text(encoding = "utf-8")
assert "GPU not ROCm-accessible" in source
if __name__ == "__main__":
pytest.main([__file__, "-v"])