Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

What is the training config? #6

Open
mkw18 opened this issue Dec 27, 2022 · 3 comments
Open

What is the training config? #6

mkw18 opened this issue Dec 27, 2022 · 3 comments

Comments

@mkw18
Copy link

mkw18 commented Dec 27, 2022

Hello, thanks for your work! I want to try to implement this work myself, but I cann't achieve the high performance by xP3 and mT0-xxl as shown in the paper Crosslingual Generalization through Multitask Finetuning. I wonder the training details of this work, how many steps do you train the model, and what is your lr-decay-ratio? Could I get the config file to implement your result? Thank you very much!

@Muennighoff
Copy link
Collaborator

I've reached out to @adarob, who knows those details & has the config files - Will let you know if we can release them!

@adarob
Copy link

adarob commented Dec 27, 2022

It's just the default T5X finetune configuration (https://github.com/google-research/t5x/blob/main/t5x/configs/runs/finetune.gin) with the following overrides:

BATCH_SIZE = 512
LOSS_NORMALIZING_FACTOR = 'AVERAGE_PER_SEQUENCE'
TASK_FEATURE_LENGTHS = {'inputs': 1024, 'targets': 1024}
train/utils.DatasetConfig.pack = False

Of course, you'd also need to set up the mixture if you're using SeqIO.

We only trained for 30k steps and picked the best checkpoint, which I believe was around 7k.

@mkw18
Copy link
Author

mkw18 commented Dec 28, 2022

Thanks a lot! I will try this.

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

No branches or pull requests

3 participants