Skip to content

fix: resuming failed because wct was reinitialized #1070

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

Closed
hawkoli1987 wants to merge 1 commit into from
Closed

fix: resuming failed because wct was reinitialized #1070

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
8 changes: 8 additions & 0 deletions src/megatron/bridge/training/utils/train_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -862,6 +862,14 @@ def report_throughput(
elapsed_samples = int(history_samples[-1]) - int(history_samples[0])
elapsed_tokens = int(history_tokens[-1]) - int(history_tokens[0])
elapsed_wct = history_wct[-1] - history_wct[0]

# Skip throughput calculation if elapsed_wct is zero or negative
# This can happen during checkpoint resumption when history_wct is reinitialized
# and the first few iterations are very fast or have identical timestamps
if elapsed_wct <= 0:
print_rank_0(f"Warning: elapsed_wct is {elapsed_wct}, skipping throughput calculation at iteration {iteration}")
return {}

batches_per_sec = elapsed_batches / elapsed_wct
samples_per_sec = elapsed_samples / elapsed_wct
dev_batches_per_sec = batches_per_sec / world_size
Expand Down
Loading