waifu-collective
/
toolbox
2
1
Fork 0
Code Issues 2 Pull Requests Releases Wiki Activity

fix: tensorboard getting messed up after epoch end

This commit is contained in:
11b 2022-12-18 22:24:26 -03:00
parent fb3dc4ba22
commit 25ae9da046
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
training/colossalai/run_clm.py
View File

@ -577,8 +577,8 @@ def main():
train_perplexity = math.exp(loss)
except OverflowError:
train_perplexity = float("inf")
writer.add_scalar("Train/Perplexity (Step)", train_perplexity, step)
writer.add_scalar("Train/Loss (Step)", loss, step)
writer.add_scalar("Train/Perplexity (Step)", train_perplexity, global_step)
writer.add_scalar("Train/Loss (Step)", loss, global_step)
if args.output_dir is not None and args.checkpointing_steps is not None:
if args.checkpointing_steps != "epoch" and completed_steps % int(args.checkpointing_steps) == 0: