linkedin / detext

DeText: A Deep Neural Text Understanding Framework for Ranking and Classification Tasks
BSD 2-Clause "Simplified" License
1.26k stars 134 forks source link

override the steps_per_eval when num_eval_rounds provided #18

Closed cyzhao2013 closed 4 years ago

cyzhao2013 commented 4 years ago

Description

In order to easily use katib hyperparameter tuning, we want keep the steps_per_eval unchanged when batch size changed.

Fixes # (issue)

Type of change

Please delete options that are not relevant.

List all changes

Please list all changes in the commit.

I set total_eval_steps in run_detext.sh. Then test it using bash run_detext.sh. The steps_per_eval will set by max(1, num_train_steps / total_eval_steps)

Test Configuration:

Checklist