[Refactor] Remove Fp8OnlineLinearMethod as scheduled (#45463)

Signed-off-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Wentao Ye
2026-06-16 04:35:58 -07:00
committed by GitHub
co-authored by mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent d53f4593ce
commit 3f53e2138f
5 changed files with 25 additions and 82 deletions
@@ -109,6 +109,10 @@ class Fp8PerTensorOnlineLinearMethod(_Fp8OnlineLinearBase):
def __init__(self):
super().__init__()
self.block_quant = False
self.use_deep_gemm = False
self.use_marlin = False
self.marlin_input_dtype = None
self.weight_quant_key = kFp8StaticTensorSym
# Use per-token quantization for better perf if dynamic and cutlass
if cutlass_fp8_supported():
@@ -144,6 +148,7 @@ class Fp8PerTensorOnlineLinearMethod(_Fp8OnlineLinearBase):
out_dtype=self.out_dtype,
module_name=self.__class__.__name__,
)
self.use_marlin = isinstance(self.fp8_linear, MarlinFP8ScaledMMLinearKernel)
def process_weights_after_loading(self, layer: Module) -> None:
if getattr(layer, "_already_called_process_weights_after_loading", False):
@@ -156,6 +161,8 @@ class Fp8PerTensorOnlineLinearMethod(_Fp8OnlineLinearBase):
replace_parameter(layer, "weight", qweight.t().data)
replace_parameter(layer, "weight_scale", weight_scale.data)
if self.use_marlin and hasattr(self.fp8_linear, "marlin_input_dtype"):
self.fp8_linear.marlin_input_dtype = self.marlin_input_dtype
self.fp8_linear.process_weights_after_loading(layer)
# Prevent duplicate processing (e.g., during weight reload)