fix muxi int8 (#535)

1. 修复muxi int8-vllm推理结果精度问题。torch.empty导致推理结果有nan值。
Co-authored-by: root <root@master.cluster.local>

lightx2v/common/ops/mm/mm_weight.py

@@ -625,7 +625,7 @@ class MMWeightWint8channelAint8channeldynamicVllm(MMWeightQuantTemplate):
         shape = (input_tensor.shape[0], self.weight.shape[1])
         dtype = input_tensor.dtype
         device = input_tensor.device
-        output_tensor = torch.empty(shape, dtype=dtype, device=device, requires_grad=False)
+        output_tensor = torch.zeros(shape, dtype=dtype, device=device, requires_grad=False)
 
         input_tensor_quant, input_tensor_scale = self.act_quant_func(input_tensor)
         torch.ops._C.cutlass_scaled_mm(