Made DDPOptimizer work with HOPs #138787
Made DDPOptimizer work with HOPs #138787
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138787
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (2 Unrelated Failures)As of commit d21fc91 with merge base 8fbf866 ( FLAKY - The following job failed but was likely due to flakiness present on trunk:
BROKEN TRUNK - The following job failed but was present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
ciflow/inductor
module: dynamo
oncall: distributed
cc H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o ezyang SherlockNoMad EikanWang jgong5 wenzhe-nrv voznesenskym penguinwu Guobing-Chen XiaobingSuper zhuhaozhe blzheng jiayisunx chenyang78 kadeng chauhang amjames rec [ghstack-poisoned]
Fixes #137481 cc H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o ezyang SherlockNoMad EikanWang jgong5 wenzhe-nrv voznesenskym penguinwu Guobing-Chen XiaobingSuper zhuhaozhe blzheng jiayisunx chenyang78 kadeng chauhang amjames rec [ghstack-poisoned]
Fixes #137481 cc H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o ezyang SherlockNoMad EikanWang jgong5 wenzhe-nrv voznesenskym penguinwu Guobing-Chen XiaobingSuper zhuhaozhe blzheng jiayisunx chenyang78 kadeng chauhang amjames rec [ghstack-poisoned]
torch/fx/passes/split_module.py
Outdated
| if orig_node.op == "get_attr": | ||
| assert isinstance(orig_node.target, str) | ||
| placeholder = partition.graph.get_attr(orig_node.target) | ||
| partition.targets[orig_node.target] = _get_attr_from_qualname( |
There was a problem hiding this comment.
i don't recall what the .targets property of a graph means. is it a list of all the things that the graph accesses via .getattr?
There was a problem hiding this comment.
oh, does this mean that we set an attribute on partition so that we can access partition.name instead of accessing parent_mod.name?
There was a problem hiding this comment.
It's just the attributes that are assigned to the graphmodule. So it's a Dict from attribute name to value.
Fixes #137481 cc H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o ezyang SherlockNoMad EikanWang jgong5 wenzhe-nrv voznesenskym penguinwu Guobing-Chen XiaobingSuper zhuhaozhe blzheng jiayisunx chenyang78 kadeng chauhang amjames rec [ghstack-poisoned]
Chillee
requested review from
angelayi,
avikchaudhuri,
tugsbayasgalan,
ydwu4 and
zhxchen17
as code owners
|
@pytorchbot merge -f "queue stuck" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
5 participants
Stack from ghstack (oldest at bottom):
Fixes #137481
cc @H-Huang @awgu @kwen2501 @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @c-p-i-o @ezyang @SherlockNoMad @EikanWang @jgong5 @wenzhe-nrv @voznesenskym @penguinwu @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @jiayisunx @chenyang78 @kadeng @chauhang @amjames @rec