Casrel论文代码遇到的问题，已解决

1.UnicodeDecodeError: ‘gbk’ codec can’t decode byte 0xae in position 149: illegal multibyte sequence

将with open(file) as f: 改成 with open(file, ‘r’, encoding=‘utf-8’) as f:


1
2
3

2、AttributeError: module ‘tensorflow’ has no attribute ‘ConfigProto’

这是由于tensorflow 2.0版本与1.0版本的用法不兼容导致的错误解决方法
将tf.ConfigProto改为tf.compat.v1.ConfigProto
1
2

3、.train计算损失时 print(‘>> epoch:’, e, ‘batch:’, b, ‘loss:’, loss.item())
一直出现RuntimeError: expected backend CPU and dtype Float but got backend CPU and dtype Long

源代码
train：loss = model.loss_fn(true_y, pred_y, input_mask)
model：return torch.sum(loss * mask) / torch.sum(mask)
mask = mask.unsqueeze(-1)
1
2
3
4

经过打印发现只有mask不是float类型
修改后

train：loss = model.