[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Harry Mellor
2026-09-16 10:57:24 -07:00
committed by GitHub
co-authored by Claude Opus 5
parent 1085b64425
commit 651a88c09a
693 changed files with 2088 additions and 4171 deletions
@@ -881,8 +881,7 @@ def choose_mp_linear_kernel(
def init_mxfp8_linear_kernel(*, bmm_batch_size: int | None = None) -> Mxfp8LinearKernel:
"""Select and instantiate the best MXFP8 linear kernel for the
current platform.
"""
current platform."""
config = Mxfp8LinearLayerConfig(bmm_batch_size=bmm_batch_size)
platform = current_platform._enum
@@ -934,8 +933,7 @@ def init_mxfp4_linear_kernel(
activation_quant_key: QuantKey | None = None,
) -> MxFp4LinearKernel:
"""Select and instantiate the best MXFP4 linear kernel for the
current platform.
"""
current platform."""
config = MxFp4LinearLayerConfig(
activation_quant_key=activation_quant_key,
)
@@ -982,8 +980,7 @@ def init_mxfp6_linear_kernel(
activation_quant_key: QuantKey | None = None,
) -> MxFp6LinearKernel:
"""Select and instantiate the best MXFP6 linear kernel for the
current platform.
"""
current platform."""
config = MxFp6LinearLayerConfig(
weight_quant_key=weight_quant_key,
activation_quant_key=activation_quant_key,
@@ -1071,8 +1068,7 @@ def init_wfp8_a16_linear_kernel(
def init_nvfp4_linear_kernel(use_a16: bool = False) -> NvFp4LinearKernel:
"""Select and instantiate the best NVFP4 linear kernel for the
current platform.
"""
current platform."""
config = NvFp4LinearLayerConfig()
a16_kernels = (
FlashInferCuteDslNvFp4W4A16LinearKernel,
@@ -177,8 +177,7 @@ class FlashInferCutlassNvFp4LinearKernel(NvFp4LinearKernel):
def input_quant_key(self) -> QuantKey | None:
"""This kernel supports dynamic quantization of the input. By
convention, pre-quantized blockscales must use the swizzled layout.
"""
convention, pre-quantized blockscales must use the swizzled layout."""
return kNvfp4Dynamic
@classmethod
@@ -41,8 +41,7 @@ class FP8BlockParams(FP8Params):
@property
def block_scale_attr(self) -> str:
"""Fp8LinearMethod registers the block scale as ``weight_scale_inv``,
compressed-tensors as ``weight_scale``.
"""
compressed-tensors as ``weight_scale``."""
return (
self.WEIGHT_SCALE
if self.weight_scale_inv is None
@@ -174,8 +174,7 @@ class CutlassFP8ScaledMMLinearKernel(FP8ScaledMMLinearKernel):
def input_quant_key(self) -> QuantKey | None:
"""Only static per-tensor activation quantization is supported for external
quantization.
"""
quantization."""
if self.config.activation_quant_key == kFp8StaticTensorSym:
return kFp8StaticTensorSym
return None
@@ -185,8 +184,7 @@ class CutlassFP8ScaledMMLinearKernel(FP8ScaledMMLinearKernel):
x: torch.Tensor, dim: int, alignment: int, value: float = 0.0
) -> torch.Tensor:
"""Pad tensor ``x`` along ``dim`` to the next multiple of
``alignment``.
"""
``alignment``."""
remainder = x.shape[dim] % alignment
if remainder == 0:
return x