[asynctp] Async_tp pass and ops fork + changes; Solver addition to incentivize async_tp fusable redistributions #151
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
meta-cla
bot
added
the
CLA Signed
IvanKobzarev
changed the title
IvanKobzarev
force-pushed
the
async_tp
branch
2 times, most recently
from
5d54740 to
1d4a56d
Compare
IvanKobzarev
changed the title
IvanKobzarev
force-pushed
the
async_tp
branch
5 times, most recently
from
8f8ddd5 to
ef3a9a9
Compare
IvanKobzarev
force-pushed
the
async_tp
branch
2 times, most recently
from
b981eec to
c46a0f6
Compare
incentivize asynctp
IvanKobzarev
force-pushed
the
async_tp
branch
from
c46a0f6 to
11649aa
Compare
Contributor
|
Shared this on chat. Let me go over this carefully once I'm back from PTO. IMO having "async-tp" as an option inside AutoParallel seems counterintuitive to me, but let me look into this carefully next week. |
Contributor
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Outdated suggestions cannot be applied.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.
torch/_inductor/fx_passes/micro_pipeline_tp.py,torch/distributed/_symmetric_memory/__init__.pyfor fast experimentation.PRs with changes on the top of base version to PyTorch repo:
pytorch/pytorch#162794
pytorch/pytorch#163068
pytorch/pytorch#163069
2.1 matmul + reduce_scatter (Partial -> Shard(dim), dim is not the last of matmul
(the last dim also supported, but has additional restride .contiguous() inside)
2.2 ag + matmul Shard(dim) -> R for argument_A of matmul, dim is not the last dim that will be reduced)