Skip to content

tokens_per_batch fixed to take into account DP and micro-batch accumu… #13

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.

Sign up for GitHub

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

Open
amithrm wants to merge 1 commit into
base: 70b_drop2_p1
Choose a base branch
from
Open

tokens_per_batch fixed to take into account DP and micro-batch accumu… #13

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion axlearn/experiments/text/gpt/fuji.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -101,7 +101,7 @@ def get_trainer_kwargs(
) -> Dict[str, Any]:
"""Construct default trainer kwargs given a model size."""
# tokens_per_batch = 4 * (1024**2) # 4M tokens.
tokens_per_batch = MAX_SEQUENCE_LENGTH[version] # one sequence per batch
tokens_per_batch = MAX_SEQUENCE_LENGTH[version] * int((jax.device_count()/TRN_MODEL_AXIS_SIZE)*GRADIENT_ACCUMULATION_MICROBATCHES) # one sequence per batch
max_step = TOTAL_TOKENS[version][model_size] // tokens_per_batch
max_sequence_length = MAX_SEQUENCE_LENGTH[version]
train_batch_size = tokens_per_batch // max_sequence_length
Expand Down