Skip to content

vllm.model_executor.layers.quantization.kernels.mixed_precision

Modules:

Name Description
MPLinearKernel
allspark
bitblas
conch
cutlass
dynamic_4bit
exllama
machete
marlin

_POSSIBLE_KERNELS module-attribute 

_POSSIBLE_KERNELS: list[type[MPLinearKernel]] = [
    CutlassW4A8LinearKernel,
    MacheteLinearKernel,
    AllSparkLinearKernel,
    MarlinLinearKernel,
    Dynamic4bitLinearKernel,
    BitBLASLinearKernel,
    ConchLinearKernel,
    ExllamaLinearKernel,
]

choose_mp_linear_kernel 

choose_mp_linear_kernel(
    config: MPLinearLayerConfig,
    compute_capability: int | None = None,
) -> type[MPLinearKernel]

Choose an MPLinearKernel that can implement the given config for the given compute capability. Attempts to choose the best kernel in terms of performance.

Parameters:

Name Type Description Default
config MPLinearLayerConfig

Description of the linear layer to be implemented.

 required
compute_capability Optional[int]

The compute capability of the target device, if None uses current_platform to get the compute capability. Defaults to None.

 None

Raises:

Type Description
ValueError

If no kernel can implement the given config.

Returns:

Type Description
type[MPLinearKernel]

type[MPLinearKernel]: Chosen kernel.
Source code in vllm/model_executor/layers/quantization/kernels/mixed_precision/__init__.py 
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
def choose_mp_linear_kernel(
    config: MPLinearLayerConfig, compute_capability: int | None = None
) -> type[MPLinearKernel]:
    """
    Choose an MPLinearKernel that can implement the given config for the given
     compute capability. Attempts to choose the best kernel in terms of
     performance.

    Args:
        config (MPLinearLayerConfig): Description of the linear layer to be
            implemented.
        compute_capability (Optional[int], optional): The compute capability of
            the target device, if None uses `current_platform` to get
            the compute capability. Defaults to None.

    Raises:
        ValueError: If no kernel can implement the given config.

    Returns:
        type[MPLinearKernel]: Chosen kernel.
    """
    if compute_capability is None:
        if current_platform is None:
            raise ValueError("Cannot determine compute capability")
        _cc = current_platform.get_device_capability()
        if _cc is not None:
            compute_capability = _cc[0] * 10 + _cc[1]

    failure_reasons = []
    for kernel in _POSSIBLE_KERNELS:
        if kernel.__name__ in envs.VLLM_DISABLED_KERNELS:
            failure_reasons.append(
                f" {kernel.__name__} disabled by environment variable"
            )
            continue
        if (
            compute_capability is not None
            and kernel.get_min_capability() > compute_capability
        ):
            failure_reasons.append(
                f"{kernel.__name__} requires capability "
                f"{kernel.get_min_capability()}, current compute "
                f" capability is {compute_capability}"
            )
            continue

        can_implement, failure_reason = kernel.can_implement(config)
        if can_implement:
            return kernel
        else:
            failure_reasons.append(
                f" {kernel.__name__} cannot implement due to: {failure_reason}"
            )

    raise ValueError(
        "Failed to find a kernel that can implement the "
        "WNA16 linear layer. Reasons: \n" + "\n".join(failure_reasons)
    )