-
Notifications
You must be signed in to change notification settings - Fork 236
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
为啥我这evaluate perplexity is nan #82
Comments
evaluate 那里的write_prediction self._get_feed_dict(img, formula=formula) 没啥不设置 lr 默认dropout是1啊,我这算出来ce_words_eval是nan |
我和你差不多,只有EditDistance 是0.86,跑了10h了还这样,请问你现在知道怎么解决了吗 |
|
非常感谢,我这就试试 |
他这还有一个问题,就是比如Epoch 42/600实际上训练的是Epoch41 保存的也是Epoch40,也就是说你看他显示的开始了Epoch 42,实际上开始的是Epoch 41, 然后保存的是Epoch 40
你用的是full 数据 还是 small 数据啊?得用full数据 |
我用的是我们老师给的数据)hh |
哈喽!我能问下你们用的python和TensorFlow是什么版本的吗? |
我想更新一下,就是不要使用我上面说的在evaluate带入dropout lr的办法,这样就不是evaluate了,然是training了。还有perplexity is nan的原因可能有多种,这个项目是使用tensorflow1.X写的,目前1.X都已经无法下载了,因此需要兼容到tensorflow2.X 上,需要大改许多代码(比较麻烦,因为没改到对的地方可能会导致训练预期结果不对)。python3.9 3.10 3.11都可以,(注tensorflow最后一个支持Windows 显卡训练的版本是2.10),当然没有7~8G显存就不要考虑用显卡训练了。 |
Eval: BLEU-4 is 0.00 || ExactMatchScore is 0.00 || EditDistance is 0.00 || perplexity is 0nan
The text was updated successfully, but these errors were encountered: