[Pipelining] Allow non-0 stages to accept kwargs #136416
Closed
Conversation
This file contains hidden or 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
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/136416
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 9d17b53 with merge base 3bc073d ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
the
oncall: distributed
kwen2501
added
release notes: distributed (pipeline)
kwen2501
requested review from
H-Huang and
wconstab
and removed request for
H-Huang
For supporting usage case in torchchat:
all non-0 stages requires `input_pos` and `cache_lane`.
```
kwargs = {"input_pos": input_pos, "cache_lane": lane}
if pp_rank == first_pp_rank:
output = decorder.step(new_token, **kwargs)
elif pp_rank == last_pp_rank:
output = decorder.step(**kwargs)
else: # middle pp ranks
decorder.step(**kwargs)
```
The `forward_one_chunk` code today hard sets `{}` as kwarg for non-0 stages, hence cannot support the above use case.
cc XilunWu H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o
[ghstack-poisoned]
For supporting usage case in torchchat:
all non-0 stages requires `input_pos` and `cache_lane`.
```
kwargs = {"input_pos": input_pos, "cache_lane": lane}
if pp_rank == first_pp_rank:
output = decorder.step(new_token, **kwargs)
elif pp_rank == last_pp_rank:
output = decorder.step(**kwargs)
else: # middle pp ranks
decorder.step(**kwargs)
```
The `forward_one_chunk` code today hard sets `{}` as kwarg for non-0 stages, hence cannot support the above use case.
cc XilunWu H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o
[ghstack-poisoned]
wconstab
reviewed
Sep 23, 2024
|
|
||
| if self.rank == 0: | ||
| with self.assertRaisesRegex(PipeliningShapeError, "shape mismatch"): | ||
| _run_step(torch.randn(batch_size + 1, d_hid, device=self.device)) |
There was a problem hiding this comment.
how come the shape checking got disabled even for args? i though from reading the stage code change it would only change the validation of kwargs
There was a problem hiding this comment.
I disabled the check on args in the library code when kwargs are present.
So I have to remove the assertRaisesRegex test here (otherwise it catches nothing).
This is the comment in library code:
# TODO- need a mapping of kwarg to position in self.args_recv_info
# Without it, we are not 100% sure how to match the args and
# expected_args.
wconstab
approved these changes
Sep 23, 2024
|
Yeah, this is a "fast food" PR. Thanks for the review. |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
BoyuanFeng
pushed a commit
to BoyuanFeng/pytorch
that referenced
this pull request
Sep 25, 2024
For supporting usage case in torchchat:
all non-0 stages requires `input_pos` and `cache_lane`.
```
kwargs = {"input_pos": input_pos, "cache_lane": lane}
if pp_rank == first_pp_rank:
output = decorder.step(new_token, **kwargs)
elif pp_rank == last_pp_rank:
output = decorder.step(**kwargs)
else: # middle pp ranks
decorder.step(**kwargs)
```
The `forward_one_chunk` code today hard sets `{}` as kwarg for non-0 stages, hence cannot support the above use case.
Pull Request resolved: pytorch#136416
Approved by: https://github.com/wconstab
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Stack from ghstack (oldest at bottom):
For supporting usage case in torchchat:
all non-0 stages requires
input_posandcache_lane.The
forward_one_chunkcode today hard sets{}as kwarg for non-0 stages, hence cannot support the above use case.cc @XilunWu @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @c-p-i-o