Fix SAC BC breaking and renaming to ac_freq #397
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
as titled, SAC moved to a different public API, move to the new API to avoid CI breaking
facebook-github-bot
added
the
CLA Signed
tianyu-l
approved these changes
Jun 13, 2024
soulitzer
reviewed
Jun 13, 2024
| @@ -45,6 +45,7 @@ | |||
| # currently selective per op and per layer checkpointing are supported | |||
| def checkpoint_wrapper(module, config): | |||
| if config.mode == "selective" and config.selective_ac_option == "op": | |||
| from torch.utils.checkpoint import create_selective_checkpoint_contexts | |||
There was a problem hiding this comment.
Hey, sorry for bc-breaking! I've updated the top of the PR description to have a list of things to restore prior behavior -
Two updates below that are still needed for your case:
- return a CheckpointPolicy Enum instead of bool
- accept a
ctx: SelectiveCheckpointinstead ofmode: str
def _custom_policy(ctx, func, *args, **kwargs):
mode = "recompute" if ctx.is_recompute else "forward"
mm_count_key = f"{mode}_mm_count"
if func == torch.ops.aten.mm.default:
meta[mm_count_key] += 1
# Saves output of all compute ops, except every second mm
to_save = func in no_recompute_list and not (
func == torch.ops.aten.mm.default and meta[mm_count_key] % 2 == 0
)
return CheckpointPolicy.MUST_SAVE if to_save else CheckpointPolicy.PREFER_RECOMPUTE
wanchaol
added a commit
that referenced
this pull request
Jun 13, 2024
wanchaol
added a commit
that referenced
this pull request
Jun 13, 2024
yifuwang
reviewed
Jun 13, 2024
| @@ -60,7 +61,7 @@ def _custom_policy(mode, func, *args, **kwargs): | |||
|
|
|||
| def selective_checkpointing_context_fn(): | |||
| meta = defaultdict(int) | |||
| return _pt2_selective_checkpoint_context_fn_gen(_get_custom_policy(meta)) | |||
tianyu-l
pushed a commit
to tianyu-l/torchtitan_intern24
that referenced
this pull request
Aug 16, 2024
as titled, SAC moved to a different public API, move to the new API to avoid CI breaking
tianyu-l
pushed a commit
to tianyu-l/torchtitan_intern24
that referenced
this pull request
Aug 16, 2024
tianyu-l
pushed a commit
that referenced
this pull request
Aug 16, 2024
philippguevorguian
pushed a commit
to YerevaNN/YNNtitan
that referenced
this pull request
Aug 17, 2024
as titled, SAC moved to a different public API, move to the new API to avoid CI breaking
philippguevorguian
pushed a commit
to YerevaNN/YNNtitan
that referenced
this pull request
Aug 17, 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.
as titled, SAC moved to a different public API, move to the new API to avoid CI breaking