[Misc] Fix incorrect layer type annotation in Fp8LinearMethod (#46818)

Signed-off-by: shaojinjie.sjj <shaojinjiesjj@gmail.com>
Co-authored-by: shaojinjie.sjj <shaojinjiesjj@gmail.com>
This commit is contained in:
jj shao
2026-06-27 20:23:59 +00:00
committed by GitHub
co-authored by shaojinjie.sjj
parent 56aa067bf0
commit ea2ead1db3
@@ -320,7 +320,7 @@ class Fp8LinearMethod(LinearMethodBase):
def create_weights(
self,
layer: RoutedExperts,
layer: torch.nn.Module,
input_size_per_partition: int,
output_partition_sizes: list[int],
input_size: int,
@@ -394,7 +394,7 @@ class Fp8LinearMethod(LinearMethodBase):
self.use_marlin = isinstance(self.fp8_linear, MarlinFP8ScaledMMLinearKernel)
def process_weights_after_loading(self, layer: RoutedExperts) -> None:
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
if self.use_marlin:
if not self.block_quant:
# Canonicalize to (K, N) for the kernel.