exo-explore/exo · error · ValueError

Unsupported mlx dtype for wire: {t.dtype}

Error message

Unsupported mlx dtype for wire: {t.dtype}

What it means

Error "Unsupported mlx dtype for wire: {t.dtype}" thrown in exo-explore/exo.

Source

Thrown at src/exo/worker/engines/mlx/disaggregated/adapter.py:65

        if keys.dtype in _MX_TO_STR:
            return _MX_TO_STR[keys.dtype]
        break
    return "bfloat16"


def str_to_mx_dtype(dtype: DType) -> mx.Dtype:
    if dtype not in _STR_TO_MX:
        raise ValueError(f"Unsupported wire dtype: {dtype!r}")
    return _STR_TO_MX[dtype]


def array_to_bytes(t: mx.array) -> bytes:
    # bf16 has no native numpy dtype; bitcast through uint16.
    if t.dtype == mx.bfloat16:
        return np.asarray(t.view(mx.uint16)).tobytes()
    if t.dtype in (mx.float16, mx.float32):
        return np.asarray(t).tobytes()
    raise ValueError(f"Unsupported mlx dtype for wire: {t.dtype}")


def bytes_to_array(data: bytes, shape: tuple[int, ...], dtype: DType) -> mx.array:
    match dtype:
        case "bfloat16":
            arr = np.frombuffer(data, dtype=np.uint16).reshape(shape).copy()
            return mx.array(arr).view(mx.bfloat16)
        case "float16":
            arr = np.frombuffer(data, dtype=np.float16).reshape(shape).copy()
            return mx.array(arr)
        case "float32":
            arr = np.frombuffer(data, dtype=np.float32).reshape(shape).copy()
            return mx.array(arr)


def bhsd_to_nhd(t: mx.array) -> mx.array:
    if t.ndim != 4 or int(t.shape[0]) != 1:
        raise ValueError(f"Expected BHSD with B=1, got shape={tuple(t.shape)}")

View on GitHub (pinned to 21a54c5ea0)

When it happens

Trigger: Thrown at src/exo/worker/engines/mlx/disaggregated/adapter.py:65 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of exo-explore/exo@21a54c5ea0 (2026-08-26). Data as JSON: /api/errors/0a3163224326d5ba. Report an issue: GitHub.