fix: use rolling window for mfu calculation #10
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
samsja
changed the title
samsja
changed the title
|
mfu seems correct, it start high tho and then reduce, I don't know why: https://wandb.ai/primeintellect/debug_1B_zero_band/runs/c403bo2n?nw=nwusersamsja_pi I also checked the reported tokens_per_sec is 62k which exactly matched the total_tokens = 5502402560 divided by the time of training 147.75 * 60 seconds. So I would say this pr is ready to merge very confident in the reported mfu |
JohannesHa
approved these changes
Sep 26, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
this pr fix the weird unstable mfu. It use a rolling average of tokens_per_second. The old implementation that was measuring the time per steps was unstable, lead to weird mfu (>100 with grad_acc=1).
[h100 sxm]
1b with 4 gpus : 48.70
1b with 8 gpus: 47.68