vdigpku / iternet Goto Github PK

Hello. Thank you for your great work. I try to training ITerNet in VinText dataset with Vietnamese characters such as ("Ạ", "Â", "Ơ",...). Firstly, I train the language model using my own vietnamese token but got nan loss. Can you help me to fix this ?

As mentioned in the paper, you use 20% training data(around 16M*0.2 = 3.2M) to train the model. I have some questions about it.
Previously, the baseline model ABINet consists of three stages: vision pre-training, language-pre-training and final-training.
In your settings, you use 20% data for final-training. Also, there is no doubt the BCN language model is pre-trained on WikiText.
But How much data has been utilized for vision pre-training? 100% training data or 20% data?
Hope to receive your answer, Thanks!

请问训练中文数据集可行不？可以具体讲讲操作嘛？非常感谢！

eval loss = 2.528, ccr = 0.595, cwr = 0.125, ted = 16603.000, ned = 5534, ted/w = 1.216
你好，我用Evaluation的命令测试测试集，为什么得到的这几个参数没有说明是什么？我不明白这几个参数的意义。为什么没有accuracy这个指标？

@VDIGPKU waiting code release