search

kakaobrain / g2pm

A Neural Grapheme-to-Phoneme Conversion Package for Mandarin Chinese Based on a New Open Benchmark Dataset
Apache License 2.0
336 stars 73 forks source link

Can you provide the complete code for training? #7

Open zzy1026 opened 4 years ago

zzy1026 commented 4 years ago

Hi, Thanks for the good job. I used chinese bert to do this work with your dataset, but I could't get good result like yours. So I want to study your code and know how to training model using this datasets. But the codes your offerring now are only to predict. So can you provide the complete code about your model or bert in your papper? Thank you for your help.

seanie12 commented 4 years ago

Hi, in my experiment the performance of bert is really sensitive to the choice of optimizer and learning rate. I will upload the scripts for training bert as soon as possible.

zzy1026 commented 4 years ago

Hi, in my experiment the performance of bert is really sensitive to the choice of optimizer and learning rate. I will upload the scripts for training bert as soon as possible.

Thank you! I also want to know that in your bert, the output of fc layer is the number of this polyphonic character's possible pinyin classes or all polyphonic characters?

seanie12 commented 4 years ago

Yes the number of output of fc layer is the number of all possible pinyins for polyphonic character.

Here is the code for training bert.