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

The number of training steps in the SHP dataset #73

Open
bonin147 opened this issue Mar 16, 2024 · 0 comments
Open

The number of training steps in the SHP dataset #73

bonin147 opened this issue Mar 16, 2024 · 0 comments

Comments

@bonin147
Copy link

In the complete example you provided, a training process of 160,000 steps was conducted using the HH dataset. However, when I trained using the SHP dataset, only 32,500 steps were completed, despite the SHP dataset having twice the size of the training dataset compared to the HH dataset. What could be the reason for this difference?

my code : python -u train.py model=pythia28 datasets=[shp] loss=sft exp_name=shp_sft gradient_accumulation_steps=4 batch_size=24 eval_batch_size=24 trainer=FSDPTrainer sample_during_eval=false model.fsdp_policy_mp=bfloat16

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

1 participant