Open mfxss opened 6 years ago
While training the evaluation accuracy was 89%. But only 14% using eval.py. I am using the same data. I found out what happened. I was using an old version of data_helper which shuffled test data for all. Thank you anyway.
If your data is embedded and labels are available, you can use the following repository: https://github.com/ms-unlimit/Transformer-Based-Machine-Learning-Framework
While training the evaluation accuracy was 89%. But only 14% using eval.py. I am using the same data. I found out what happened. I was using an old version of data_helper which shuffled test data for all. Thank you anyway.