[XPU] fix MoE triton backend in online fp8 quantization (#40109)
Signed-off-by: Yan Ma <yan.ma@intel.com>
This commit is contained in:
@@ -1019,10 +1019,6 @@ class Fp8OnlineMoEMethod(Fp8MoEMethod):
|
||||
layer.w2_weight[expert, :, :]
|
||||
)
|
||||
|
||||
if current_platform.is_xpu():
|
||||
w13.data = w13.transpose(-1, -2).contiguous()
|
||||
w2.data = w2.transpose(-1, -2).contiguous()
|
||||
|
||||
# Shuffle weights to runtime format and setup kernel.
|
||||
self._setup_kernel(
|
||||
layer,
|
||||
|
||||
Reference in New Issue
Block a user