Skip to content

GitLab

"vscode:/vscode.git/clone" did not exist on "f9ba3cdc8185bfe04e12e6a6ce8b100c3374cdbc"
Unverified Commit 473ff985 authored by Shaden Smith's avatar Shaden Smith Committed by GitHub
Browse files

scales throughput by logging freq (#408)

parent 4ac9bf60
Show whitespace changes
Inline Side-by-side
Showing with 2 additions and 3 deletions
deepspeed/runtime/pipe/engine.py
View file @ 473ff985
...@@ -260,9 +260,8 @@ class PipelineEngine(DeepSpeedEngine): ...@@ -260,9 +260,8 @@ class PipelineEngine(DeepSpeedEngine):
self.module.train() self.module.train()
self.total_loss = None self.total_loss = None
self.timers('train_batch').start()
# Do the work # Do the work
self.timers('train_batch').start()
sched = schedule.TrainSchedule(micro_batches=self.micro_batches, sched = schedule.TrainSchedule(micro_batches=self.micro_batches,
stages=self.num_stages, stages=self.num_stages,
stage_id=self.stage_id) stage_id=self.stage_id)
...@@ -274,7 +273,7 @@ class PipelineEngine(DeepSpeedEngine): ...@@ -274,7 +273,7 @@ class PipelineEngine(DeepSpeedEngine):
if self.global_rank == 0: if self.global_rank == 0:
elapsed = self.timers('train_batch').elapsed(reset=True) elapsed = self.timers('train_batch').elapsed(reset=True)
iter_time = elapsed / self.steps_per_print() iter_time = elapsed / self.steps_per_print()
tput = self.train_batch_size() / elapsed tput = self.train_batch_size() / iter_time
print(f'steps: {self.global_steps} ' print(f'steps: {self.global_steps} '
f'loss: {self.agg_train_loss:0.4f} ' f'loss: {self.agg_train_loss:0.4f} '
f'iter time (s): {iter_time:0.3f} ' f'iter time (s): {iter_time:0.3f} '
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment