Dive-into-DL-PyTorch icon indicating copy to clipboard operation
Dive-into-DL-PyTorch copied to clipboard

Published 20 hours ago • verified publisher icon ShusenTang

6.5.2 RNN简洁实现,使用相邻采样时,state在新的epoch中未被初始化

Open jianli-Alex opened this issue 3 years ago • 0 comments

bug描述 2020-11-06 16-43-35 的屏幕截图 state = None放在了for epoch in range(num_epochs)上. 因此,在新的epoch中,state实际上没有被初始化,而是使用上一epoch最后一次迭代最后一个时间步的state, 这样并不合理. 要把state = None放在了for epoch in range(num_epochs)下面,for X, Y in data_iter上面

版本信息 pytorch: 1.6.0 torchvision: 0.7.0 torchtext: ...

jianli-Alex avatar Nov 06 '20 08:11 jianli-Alex