[Distributed] Add lanes to KV cache #1174
Merged
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchchat/1174
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 9514b54 with merge base 8d01d9b ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
kwen2501
force-pushed
the
cache_lanes
branch
from
e33e681 to
39eff90
Compare
kwen2501
changed the title
lessw2020
reviewed
Sep 23, 2024
| @@ -445,11 +456,12 @@ def get_example_ins_outs(seqlen: int) -> Tuple[torch.Tensor, torch.Tensor]: | |||
| group=pp_group, | |||
| ) | |||
| # create schedule | |||
| decode_schedule = ScheduleGPipe(decode_stage, mbs) | |||
| decorder = ScheduleGPipe(decode_stage, 1) | |||
There was a problem hiding this comment.
syntax error - this should be 'decoder' and not 'decorder'.
lessw2020
reviewed
Sep 23, 2024
| @@ -467,11 +479,11 @@ def get_example_ins_outs(seqlen: int) -> Tuple[torch.Tensor, torch.Tensor]: | |||
|
|
|||
| # Run data through pipeline | |||
| if pp_rank == first_pp_rank: | |||
| output = decode_schedule.step(new_token) | |||
| output = decorder.step(new_token, **kwargs) | |||
There was a problem hiding this comment.
same, syntax error - this should be 'decoder' and not 'decorder'.
lessw2020
reviewed
Sep 23, 2024
| elif pp_rank == last_pp_rank: | ||
| output = decode_schedule.step() | ||
| output = decorder.step(**kwargs) |
There was a problem hiding this comment.
same, syntax error - this should be 'decoder' and not 'decorder'.
lessw2020
reviewed
Sep 23, 2024
| else: # middle pp ranks | ||
| decode_schedule.step() | ||
| decorder.step(**kwargs) |
There was a problem hiding this comment.
last one, syntax error - this should be 'decoder' and not 'decorder'.
lessw2020
approved these changes
Sep 23, 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.
KV cache is extended to have multiple lanes, each letting a separate batch pass through, achieving pipeline parallelism.
Major changes
setup_cacheswill take one kwargcache_lanes(default to 1).attention.kv_cacheis now ann.ModuleList, containing multipleKVCache's, each corresponding to a lane.We now pass
kwargs = {"input_pos": input_pos, "cache_lane": lane}to thestep()function. Removing the temporary helper functionmodel.setup_input_pos.Requires pytorch/pytorch#136416 to support pass-in of kwargs.