jax-ml/jax · error · BufferError

The device specification passed to to_dlpack contains an uns

Error message

The device specification passed to to_dlpack contains an unsupported device type (DLDeviceType: {dl_device_type})

What it means

The dl_device tuple passed to __dlpack__/to_dlpack names a DLPack device type (e.g. kDLCUDA, kDLCPU, kDLMetal) that JAX does not map to any supported backend platform. Per the array API spec, unsupported producers raise BufferError.

Source

Thrown at jax/_src/dlpack.py:145

    is mutated, it may lead to undefined behavior when using the associated JAX
    array. When JAX eventually supports ``DLManagedTensorVersioned``
    (DLPack 1.0), it will be possible to specify that a buffer is read-only.
  """
  if not isinstance(x, array.ArrayImpl):
    raise TypeError("Argument to to_dlpack must be a jax.Array, "
                    f"got {type(x)}")

  device = None
  dl_device_type, local_hardware_id = dl_device if dl_device else (None, None)
  if dl_device_type:
    try:
      dl_device_platform = _DL_DEVICE_TO_PLATFORM[dl_device_type]
      backend = xla_bridge.get_backend(dl_device_platform)
      device = backend.device_from_local_hardware_id(local_hardware_id)
    except KeyError:
      # https://data-apis.org/array-api/latest/API_specification/generated/array_api.array.__dlpack__.html
      # recommends using BufferError.
      raise BufferError(
          "The device specification passed to to_dlpack contains an"
          f" unsupported device type (DLDeviceType: {dl_device_type})"
      ) from None

  # As new versions are adopted over time, we can maintain some legacy paths
  # for compatibility mediated through the max_version parameter.
  # TODO(micky774): Deprecate default usage of DLPackManagedTensor when XLA
  # supports DLManagedTensorVersioned (DLPack version 1.0) and repurpose the
  # current _to_dlpack as a legacy path for (0,5) <= max_version < (1,0).
  if max_version is None or max_version >= DLPACK_VERSION:
    # Latest
    return _to_dlpack(
      x, stream=stream,
      src_device=src_device,
      device=device,
      copy=copy
    )
  elif max_version >= MIN_DLPACK_VERSION:

View on GitHub (pinned to 1e1c6a8fc0)

Solutions

1. Request a supported device type: CPU or the GPU type JAX was built for (kDLCUDA, or kDLCUDAManaged where supported)
2. Pass dl_device=None to let JAX choose the array's native device
3. Verify the target platform backend is installed (jax.cuda / jax.rocm)

Example fix

# before
jax_arr.__dlpack__(dl_device=(DLDeviceType.kDLOpenCL, 0))

# after
jax_arr.__dlpack__()  # or dl_device=(DLDeviceType.kDLCUDA, 0)

from jax._src.dlpack import _DL_DEVICE_TO_PLATFORM
if dl_device and dl_device[0] not in _DL_DEVICE_TO_PLATFORM:
    dl_device = None  # fall back to native device

try:
    x.__dlpack__(dl_device=dl_device)
except BufferError:
    x.__dlpack__()

When it happens

Trigger: Calling x.__dlpack__(dl_device=(t, id)) with a DLDeviceType enum value not in JAX's _DL_DEVICE_TO_PLATFORM mapping (e.g. kDLROCM on a build without ROCm, kDLOpenCL, kDLVulkan).

Common situations: PyTorch requesting an OpenCL/Vulkan/Metal device buffer; consumers hard-coding a device enum not supported by the installed JAX backends.

Related errors

• __dlpack__ only supported for unsharded arrays.
• The array passed to from_dlpack must have __dlpack__ and __d
• Array passed to from_dlpack is on unsupported device type (D
• to_dlpack can only pack a dlpack tensor from an array on a s
• Unknown GPU platform for __dlpack__: {platform_version}

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/eae0f1b7a22d6520. Report an issue: GitHub.