原因是mask错误,被mask的部分应该为True
data = [1, 2, 3]
len = 5
mask = [False, False, False, True, True]
ref: My transformer NMT model is giving "nan" loss value - #6 by FeryET - nlp - PyTorch Forums
原因是mask错误,被mask的部分应该为True
data = [1, 2, 3]
len = 5
mask = [False, False, False, True, True]
ref: My transformer NMT model is giving "nan" loss value - #6 by FeryET - nlp - PyTorch Forums