[FP8]add FP8 WoQ kernel abstraction. (#32929)
Signed-off-by: Zhu, Zufang <zufang.zhu@intel.com>
This commit is contained in:
@@ -22,7 +22,6 @@ from vllm.model_executor.layers.quantization.base_config import (
|
||||
QuantizeMethodBase,
|
||||
)
|
||||
from vllm.model_executor.layers.quantization.utils.marlin_utils_fp8 import (
|
||||
apply_fp8_marlin_linear,
|
||||
prepare_fp8_layer_for_marlin,
|
||||
)
|
||||
from vllm.model_executor.layers.quantization.utils.quant_utils import (
|
||||
@@ -177,15 +176,4 @@ class FBGEMMFp8LinearMethod(LinearMethodBase):
|
||||
x: torch.Tensor,
|
||||
bias: torch.Tensor | None = None,
|
||||
) -> torch.Tensor:
|
||||
if self.quant_config.use_marlin:
|
||||
return apply_fp8_marlin_linear(
|
||||
input=x,
|
||||
weight=layer.weight,
|
||||
weight_scale=layer.weight_scale,
|
||||
workspace=layer.workspace,
|
||||
size_n=layer.output_size_per_partition,
|
||||
size_k=layer.input_size_per_partition,
|
||||
bias=bias,
|
||||
)
|
||||
|
||||
return self.fp8_linear.apply_weights(layer, x, bias)
|
||||
|
||||
Reference in New Issue
Block a user