jhyuklee / ethnicity-tensorflow Goto Github PK

Hello There, I just read your paper, it is very informative and helpful. I just want to know that why the data_ijcai_author has been used as the testing dataset and not the file data_raw_test?

As the system is not able to read the data_ijcai_author for some reason unknown, so i wanted to ask if we can use the data_raw_test file as its replacement?

Thanks in advance

Hi there, thanks for being so helpful with my last issue.
The one I am having right now that, I don't know tensorflow yet. I am working my way through your code.

1. Its a silly question from my side, but still I would like to know that whenever you are passing n-grams to the corresponding LSTM model, then are you collecting the ouputs of all the time steps to forward em to the linear layer or only the last output of the LSTM for that?

2. The padding that you have done for making the LSTM outputs of same size for all the data points via tf.gather(which has a very interesting and a particular way of padding the input based on the max_time_step), does it needs to be padded that way only or you can we pad it in other ways as well?

I know, I am maybe a little late to the party, but hoefully you can still help me out.
I am relativly new to python, but would still highly appreciate being able to use your work.

Unfortunately after running python main.py I encounter the following error:

'''
File "main.py", line 136, in main
saved_params['checkpoint_dir'] += model_name
TypeError: unsupported operand type(s) for +=: 'Flag' and 'str'
'''

I am also not 100% sure, how to provide / specify the input data (the names for which the nationality should be predicted).

Hopefully you can help me a little.
Thank you in advance!