Confusion about some index/slice operators in your code

[Cute-Yang]

slice operator

Hello,sir,your model is so excellent which makes me learn a lot. but I have several problems about your code which make me feel confused:

• 2-D array assign row value

  you define a 2-D matrix, as v_text_embed,your code is like

  v_text_embed = np.zeros((self.args.V_n, self.args.in_f_d * 5))
  feature_temp = np.reshape(np.asarray(feature_temp), [1, -1])
  v_text_embed[i] = feature_temp

  absolutely,you assgine your array row by row,but you reshape the row to a (1,-1) like array,I think it will make people confused,however,the syntax is correct and the code can run

• torch Tensor slice

  you use some confused operator in Tensor operator,your code is like

  a_net_embed_batch = self.feature_list[6][id_batch]
  a_text_embed_batch_1 = self.feature_list[7][id_batch, :embed_d][0]
  a_text_embed_batch_2 = self.feature_list[7][id_batch, embed_d : embed_d * 2][0]
  a_text_embed_batch_3 = self.feature_list[7][id_batch, embed_d * 2 : embed_d * 3][0]

  however, all of them are 2-D tensor,and your id_batch is define as a 2-D array with shape(1,-1), your first operator is tensor[id_batch] and next operator is tensor[id_batch,:],also they looks like similar, but because of id_batch is 2-D array,the next operator will add a dim to new tensor,so you use [0] operator to reuce it if you use a 1-D list as the id_batch,the operator will not produce result which make people confused

finaly,wish you have better work and look forward to your future works!