search

tensorflow / tensor2tensor

Library of deep learning models and datasets designed to make deep learning more accessible and accelerate ML research.
Apache License 2.0
15.6k stars 3.51k forks source link

Hyper parameter tuning for local execution #1490

Open Drunkar opened 5 years ago

Drunkar commented 5 years ago

Description

I know t2t has a hyperparameter tuning functin, but it's only for ML Engine. I implented hyperparameter tuning with Optuna for t2t v1.10.0.

https://github.com/Drunkar/tensor2tensor-optuna

Opuna is a open-sourced hyperparameter optimization framework developed by Preferred Networks.

https://optuna.org/

I want to ask does someone need this feature or not because ML Engine is sometimes high-cost. Thanks.

Environment information

OS: Ubuntu 16.04.6 LTS

$ pip freeze | grep tensor
mesh-tensorflow==0.0.5
tensor2tensor==1.10.0
tensorboard==1.12.2
tensorflow==1.12.0
tensorflow-datasets==1.0.1
tensorflow-hub==0.3.0
tensorflow-metadata==0.13.0
tensorflow-probability==0.6.0
tensorflow-serving-api==1.13.0

$ python -V
Python 3.6.7 :: Anaconda custom (64-bit)
afrozenator commented 5 years ago

This looks great @Drunkar ! Please feel free to send in a PR to me and @lukaszkaiser and we'll take a look -- but we'd like to add this if possible.

Drunkar commented 5 years ago

thanks for commets! will PR after rebase it to master branch...

ghost commented 4 years ago

Hello, it looks like there is no PR yet. @afrozenator and @lukaszkaiser, will there be any update about hyper parameter tuning for local execution?