qq_43118572的博客

dim 参数对应的维数消失this_platf_target_outputs = torch.reshape(this_platf_target_outputs, shape=(self.batch_size, self.seq_len, self.num_nodes, self.units)) # (64, 6, 30, 16)this_platf_target_outputs = torch.sum(this_platf_target_outputs, dim=2) # shape=(ba.

tf.multiply : 按元素相乘 # [64*6*30, 2] * [64*6*30,1]this_platf_target_outputs = tf.multiply(neighbor_part_inputs, mask[target_node, platf, ...])Returns x * y element-wisetf.matmul: 矩阵相乘this_platf_target_outputs = tf.matmul(this_platf_target_outputs.

mask： # (30,3,64630,1)# mask[target_node,platf,...] ： ... 将后面的维数拉平# [64*6*30, 2] * [64*6*30,1]this_platf_target_outputs = tf.multiply(neighbor_part_inputs, mask[target_node,platf,...])