Open bxwldljh opened 1 year ago
Hi, nice work and I am interested in it. In train.py, there are a few lines of code I do not understand:
m1 = copy.deepcopy(sen_mask) ##[0,0,0...0,1,1,1,1] m1.scatter_(1, w_ind, 0) ##[0,0,0...0,0,1,1,0] m2 = 1 - m1 ##[1,1,1...1,1,0,0,1] if dataset=='cpv1': m3=m1*18330 else: m3 = m1 * 18455 ##[0,0,0...0,0,18455,18455,0] q2 = q2 * m2.long() + m3.long()
can you tell me what does 18330 and 18455 mean?
it is the padding_idx of word dictionary, see more in dataset.py
Hi, nice work and I am interested in it. In train.py, there are a few lines of code I do not understand:
can you tell me what does 18330 and 18455 mean?