自动进化神经网络，特征进行add时维度不匹配应该怎么改呢

Question

gpu_id: 0 , [['input', 0, 0], ['S_ResBlock_64_1', 0, 0], ['S_ConvBlock_128_1', 0, 0], ['S_ResBlock_64_1', 0, 2], ['S_ConvBlock_128_3', 1, 3], ['Max_Pool', 2, 3], ['S_ConvBlock_128_5', 3, 5], ['S_ResBlock_128_5', 4, 6], ['Max_Pool', 5, 6], ['S_ResBlock_64_1', 6, 5], ['Sum', 7, 8], ['S_ConvBlock_32_3', 10, 7], ['Sum', 9, 11], ['S_ConvBlock_32_5', 12, 8], ['S_ResBlock_128_3', 13, 13], ['full', 14, 14]]
Files already downloaded and verified
Files already downloaded and verified
train num 50000
GPUID : 0
epoch_num : 5
batch_size: 64
Total: 1527178 Trainable: 1527178
initialization method [kaiming]
epoch 1
Traceback (most recent call last):
File "/Data/home/liuqingqing/2020/813cgp-cnn-pytorch/cnn_train.py", line 204, in call
output = model(input_, None)
File "/opt/anaconda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 550, in call
result = self.forward(*input, **kwargs)
File "/Data/home/liuqingqing/2020/813cgp-cnn-pytorch/cnn_model.py", line 301, in forward
return self.main(x)
File "/Data/home/liuqingqing/2020/813cgp-cnn-pytorch/cnn_model.py", line 294, in main
outputs[nodeID] = layer(outputs[self.cgp[nodeID][1]], outputs[self.cgp[nodeID][2]])
File "/opt/anaconda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 550, in call
result = self.forward(*input, **kwargs)
File "/Data/home/liuqingqing/2020/813cgp-cnn-pytorch/cnn_model.py", line 157, in forward
out = torch.add(in_data[0], in_data[1])
RuntimeError: The size of tensor a (4) must match the size of tensor b (8) at non-singleton dimension 3

Answer 1

https://blog.csdn.net/xiaosongshine/article/details/90739233