I use the default parameters set in original code, and got the following training process. The "train_decode" and "val_decode" both are null till "epoch:0, batch:1700", but the loss is still decreasing. And I changed the output format as following. So the acc keeps as 0.0, and is there something wrong with the training process?
train_decode: [，，，的的的的的的的] | val_decode: [，，，的的的的的的的] | ground_truth: [职顺带休息网友奥斯摩]
train_decode: [，，的的的的的的的的] | val_decode: [，，的的的的的的的的] | ground_truth: [中距空对空导弹实施攻]
train_decode: [，，的的的的的的的的] | val_decode: [，，，，的的的的的的] | ground_truth: [参见第十四章“无状之]
train_decode: [，，，的的的的的的的] | val_decode: [，，，，的的的的的的] | ground_truth: [面冒个泡，得三公之教]
epoch:0, batch:19300, loss:3.55256605148, acc:0.0
train_decode: [，，，的的的的的的的] | val_decode: [，，，的的的的的的的] | ground_truth: [也许你通过跟人力资源]
train_decode: [，，，的的的的的的的] | val_decode: [，，，的的的的的的的] | ground_truth: [狼或熊则是一个好兆头]
train_decode: [，，的的的的的的的的] | val_decode: [，，，，的的的的的的] | ground_truth: [网来说是一个不小的挑]
train_decode: [，，的的的的的的的的] | val_decode: [，，，，的的的的的的] | ground_truth: [无神论科学家的说法是]
train_decode: [，，的的的的的的的的] | val_decode: [，，，，的的的的的的] | ground_truth: [闻玉帝求又破右胁而出]
epoch:0, batch:19400, loss:3.53783345222, acc:0.0
train_decode: [，，的的的的的的的的] | val_decode: [，，的的的的的的的的] | ground_truth: [机械学院→华北工学院]
train_decode: [，的的的的的的的的的] | val_decode: [，的的的的的的的的的] | ground_truth: [中。同时身体向下弯曲]
train_decode: [，的的的的的的的的的] | val_decode: [，，，，的的的的的的] | ground_truth: [到了德佩罗（Desp]
train_decode: [，，，的的的的的的的] | val_decode: [，，，的的的的的的的] | ground_truth: [这完全是一个主权国家]
train_decode: [，的的的的的的的的的] | val_decode: [，，的的的的的的的的] | ground_truth: [时候他会永远离不开老]
I use the default parameters set in original code, and got the following training process. The "train_decode" and "val_decode" both are null till "epoch:0, batch:1700", but the loss is still decreasing. And I changed the output format as following. So the acc keeps as 0.0, and is there something wrong with the training process?