search

google-research / albert

ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
Apache License 2.0
3.23k stars 570 forks source link

How to prepare for the environment/data for fine-tuning? #216

Closed 517030910405 closed 3 years ago

517030910405 commented 4 years ago

I am trying to run run_glue.sh for COLA. I downloaded the model from "https://tfhub.dev/google/albert_base/3", because the server cannot access google. However, I found that vocab.txt is not in the folder. Where should I get it? Also, I found that the program started to download even if the model albert_base_3.tar.gz is there. How to use local model? Thank you very much!

517030910405 commented 4 years ago

Is there an example of how should I organize the folders? Thanks

Maybe something like:

-ALBERT --albert_base_3 (downloaded from https://tfhub.dev/google/albert_base/3 ) ---saved_model.pb --CoLA (downloaded from https://gluebenchmark.com/tasks) ---dev.tsv ---train.tsv ---train.tsv

yhshu commented 3 years ago

I have the same question and have no idea how to organize the data. It seems nowhere mentioned the method to construct the fine-tuning data.