Skip to content

Log data-loader wait time as an input-starvation metric #564

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

Merged
jlamypoirier merged 2 commits into from
Jul 10, 2026
Merged

Log data-loader wait time as an input-starvation metric #564

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
807cc60
Log data-loader wait time as an input-starvation metric
jlamypoirier Jul 7, 2026
9e4981a
Review fixes: drop dead default, format data_wait_time_ms like siblin…
jlamypoirier Jul 10, 2026
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
5 changes: 5 additions & 0 deletions fast_llm/engine/schedule/runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -159,6 +159,9 @@ def run_step(
assert self._support_training

metrics = {} if return_metrics else None
if metrics is not None:
# Always present on logging steps so "no wait" shows as 0 rather than a gap.
metrics["data_wait_time_ms"] = 0.0
# Set the context.
context = BatchContext(
iteration=iteration,
Expand Down Expand Up @@ -433,6 +436,8 @@ def _get_forward_input(self, context: BatchContext, step: Step) -> torch.Tensor:
next(context.data_iterator)

data_time = (time.perf_counter() - start_time) * 1000
if context.metrics is not None:
context.metrics["data_wait_time_ms"] += data_time
if data_time > self._config.data_batch_warn_time_ms:
logger.warning(f"Data loading took {data_time:,.2f} ms")
return context.inputs.pop(step.global_index).detach().requires_grad_(step.stage != 0)
Expand Down
2 changes: 2 additions & 0 deletions fast_llm/logging.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,6 +65,7 @@
"skipped_iters",
"nan_iters",
"step_time_average_ms",
"data_wait_time_ms",
"remaining_time",
"completion_time",
"percent_done",
Expand All @@ -77,6 +78,7 @@
" | skipped iterations: {skipped_iters:3.0f}"
" | nan iterations: {nan_iters:3.0f}"
" | average step time {step_time_average_ms:.2f} ms"
" | data wait: {data_wait_time_ms:.2f} ms"
" | remaining {remaining_time} "
" | completion {completion_time} ({percent_done:.2f} %)"
)
Expand Down
Loading