Skip to content
This repository has been archived by the owner on May 2, 2024. It is now read-only.

How to reproduce the result reported in the table? #2

Closed
haoransh opened this issue Apr 8, 2018 · 2 comments
Closed

How to reproduce the result reported in the table? #2

haoransh opened this issue Apr 8, 2018 · 2 comments

Comments

@haoransh
Copy link

haoransh commented Apr 8, 2018

Hi, I'm trying to reproduce the result reported in the table recently. And I can only achieve ~23 with my pipeline. I'm afraid there might be something wrong.
According to your readme.md and your claim here tensorflow/tensor2tensor#611, I'm training the model with

t2t-trainer --data_dir=t2t_data --problems=translate_envi_iwslt32k --model=transformer --hparams_set=transformer_base_single_gpu --output_dir=t2t_output --max_train_steps=125000

I'm not sure whether you run the model with such hyparameters, because in your readme.md you used transformer_base hparams set.
By the way, I'm using the latest version of t2t(1.5.6), but I guess that the version won't influence a lot.

Thanks and looking forward to your reply!

@stefan-it
Copy link
Owner

@Shrshore Could you solve the problem?

@haoransh
Copy link
Author

Yes, I could obtain similar performance with transformer_base_single_gpu and 125k steps. Thanks for your efforts!

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants