rknn识别不了eltwise层

求助，我有一个网络模型，结构如图所示
转换成rknn时，提示信息：
--> Building model
W extend res1_3_7 to add will cause accuracy loss, do not extend.
W extend res2_3_14 to add will cause accuracy loss, do not extend.
W extend res2_5_19 to add will cause accuracy loss, do not extend.
W extend res3_3_26 to add will cause accuracy loss, do not extend.
W extend res3_5_31 to add will cause accuracy loss, do not extend.
W extend res3_7_36 to add will cause accuracy loss, do not extend.
W extend res3_9_41 to add will cause accuracy loss, do not extend.
W extend res4_3_48 to add will cause accuracy loss, do not extend.


我发现这几层全是eltwise层
现在转换出来的rknn模型，推理的结果是不对的，我想应该是跟这个有关，请问有解决办法吗？感谢！

自己顶一下

构建的log信息在附件，求大神帮忙看下