unslothai/unsloth · error · ValueError

Invalid gpu_ids {requested_ids}: duplicate GPU IDs are not a

Error message

Invalid gpu_ids {requested_ids}: duplicate GPU IDs are not allowed.

What it means

Raised during GPU ID validation on the Vulkan path (llama.cpp Vulkan builds) when the requested gpu_ids list contains the same Vulkan ordinal twice. Vulkan selects devices by ggml ordinal (--device VulkanN), a separate index space from CUDA/ROCm ids, so only basic well-formedness checks (duplicates, negatives) apply — there is no parent-visible or physical-count cross-check on this path.

Source

Thrown at studio/backend/utils/hardware/hardware.py:2762

) -> list[int]:
    parent_visible_spec = _get_parent_visible_gpu_spec()
    parent_visible_ids = get_parent_visible_gpu_ids()
    physical_gpu_count = get_physical_gpu_count()

    if gpu_ids is None:
        return [] if is_vulkan else parent_visible_ids

    requested_ids = list(gpu_ids)
    if len(requested_ids) == 0:
        return [] if is_vulkan else parent_visible_ids

    if is_vulkan:
        # A Vulkan build selects by ggml Vulkan ordinal (--device VulkanN), a separate
        # index space from CUDA/ROCm ids that may be empty under CPU-only torch. The
        # CUDA parent-visible / physical-count checks below do not apply; only reject
        # malformed ordinals (issue #7239).
        if len(set(requested_ids)) != len(requested_ids):
            raise ValueError(f"Invalid gpu_ids {requested_ids}: duplicate GPU IDs are not allowed.")
        negative_ids = [gpu_id for gpu_id in requested_ids if gpu_id < 0]
        if negative_ids:
            raise ValueError(
                f"Invalid gpu_ids {requested_ids}: GPU IDs must be non-negative. "
                f"Rejected IDs: {negative_ids}."
            )
        return requested_ids

    if not parent_visible_spec["supports_explicit_gpu_ids"]:
        env_var_name = (
            "ZE_AFFINITY_MASK" if get_device() == DeviceType.XPU else "CUDA_VISIBLE_DEVICES"
        )
        raise ValueError(
            f"Invalid gpu_ids {requested_ids}: explicit physical GPU IDs are "
            f"unsupported when {env_var_name} uses non-numeric or subdevice "
            f"entries ({parent_visible_spec['raw']!r}). Omit gpu_ids to use "
            "the parent-visible devices."
        )

View on GitHub (pinned to 203007d190)

Solutions

Deduplicate the gpu_ids list before passing it: sorted(set(gpu_ids)).
Fix the config/UI layer that produced the duplicated ids.
Use distinct Vulkan ordinals, one per device you want to use (e.g. [0, 1]).

Example fix

# before
apply_gpu_ids([0, 0])  # ValueError: duplicate GPU IDs are not allowed

# after
apply_gpu_ids(sorted(set([0, 0])))  # -> [0]

Defensive patterns

Strategy: validation

Validate before calling

def normalize_gpu_ids(ids):
    """Dedup while preserving order."""
    return list(dict.fromkeys(ids))

# guard: resolve_requested_gpu_ids(normalize_gpu_ids(gpu_ids))

Type guard

def are_unique_non_negative_ints(ids) -> bool:
    return (
        isinstance(ids, (list, tuple))
        and all(isinstance(i, int) for i in ids)
        and len(set(ids)) == len(ids)
        and all(i >= 0 for i in ids)
    )

Prevention

Deduplicate gpu id lists before any hardware API call.
Use set operations when merging default and user GPU selections.
Reject duplicate selections in UI multi-selects.
Remember Vulkan uses its own ordinal space — do not reuse CUDA id logic blindly.

When it happens

Trigger: Calling resolve_requested_gpu_ids([0, 0]) or any gpu_ids list with a repeated element while the active device backend is Vulkan.

Common situations: Config files where a default id list got concatenated with a user-specified one (e.g. [0] + [0, 1]); UI multi-select allowing duplicate submissions; hand-edited YAML/JSON with copy-paste duplication.

Related errors

AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15). Data as JSON: /api/errors/bea913c0e4877d5b. Report an issue: GitHub.