计算torch.nn.MultiheadAttention的FLOPS，涉及到了NonDynamicallyQuantizableLinear=0 FLOPS

[CFZ1]

pytorch版本1.10.0，每次计算到torch.nn.MultiheadAttention，都是这样汇报的：

      (attn): MultiheadAttention(
        263.17 K = 0.5576% Params, 104.86 MMACs = 0.007% MACs, 211 MFLOPS = 0.0071% FLOPs
        (out_proj): NonDynamicallyQuantizableLinear(65.79 K = 0.1394% Params, 0 MACs = 0% MACs, 0 FLOPS = 0% FLOPs, in_features=256, out_features=256, bias=True)
      )

请问这里的NonDynamicallyQuantizableLinear=0 FLOPS 没问题吗？需要自己手动计算NonDynamicallyQuantizableLinear的FLOPS，然后加上吗？谢谢