[FSDP] Fix summon_full_params test #74456
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
This test did not actually do any CPU offloading. Adding it revealed an issue in no_shard (currently only when world_size == 1) case which is tracked for a fix in #74166 Differential Revision: [D35003793](https://our.internmc.facebook.com/intern/diff/D35003793/) [ghstack-poisoned]
rohan-varma
requested review from
H-Huang,
mrshenli,
pritamdamania87 and
zhaojuanmao
as code owners
CI Flow Status⚛️ CI FlowRuleset - Version:
|
Contributor
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit e86140c (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Please report bugs/suggestions to the (internal) Dr. CI Users group. |
facebook-github-bot
added
the
oncall: distributed
This test did not actually do any CPU offloading. Adding it revealed an issue in no_shard (currently only when world_size == 1) case which is tracked for a fix in #74166 Differential Revision: [D35003793](https://our.internmc.facebook.com/intern/diff/D35003793/) [ghstack-poisoned]
This test did not actually do any CPU offloading. Adding it revealed an issue in no_shard (currently only when world_size == 1) case which is tracked for a fix in #74166 Differential Revision: [D35003793](https://our.internmc.facebook.com/intern/diff/D35003793/) [ghstack-poisoned]
18 tasks
zhaojuanmao
approved these changes
Mar 21, 2022
This test did not actually do any CPU offloading. Adding it revealed an issue in no_shard (currently only when world_size == 1) case which is tracked for a fix in #74166 Differential Revision: [D35003793](https://our.internmc.facebook.com/intern/diff/D35003793/) [ghstack-poisoned]
This test did not actually do any CPU offloading. Adding it revealed an issue in no_shard (currently only when world_size == 1) case which is tracked for a fix in #74166 Differential Revision: [D35003793](https://our.internmc.facebook.com/intern/diff/D35003793/) [ghstack-poisoned]
mrshenli
reviewed
Mar 22, 2022
Comment on lines
+41
to
+43
| lin1 = FSDP(nn.Linear(5, 5, bias=False).cuda(cls.rank), cpu_offload=cpu_offload) | ||
| lin2 = nn.Linear(5, 3, bias=False).cuda(cls.rank) | ||
| model = FSDP(nn.Sequential(lin1, lin2), cpu_offload=cpu_offload) |
Contributor
There was a problem hiding this comment.
curious, what's the expectation on the input model when cpu_offload==True? Does it have to reside on GPU? Or is it OK to pass in a CPU model?
Contributor
Author
There was a problem hiding this comment.
Passing CPU offload with CPU model should work, as we don't assume any device in the constructor and would restore to GPU in forward pass before all_gather. Although, it does not seem that we have a test for this, we should add one.
| lin2 = nn.Linear(5, 3, bias=False).cuda(cls.rank) | ||
| model = FSDP(nn.Sequential(lin1, lin2), cpu_offload=cpu_offload) | ||
| if not cpu_offload: | ||
| model = model.cuda(cls.rank) |
Contributor
There was a problem hiding this comment.
why do we need this? I assume when cpu_offload==False, the model is already on cuda, no?
Contributor
Author
There was a problem hiding this comment.
Right, we can remove this because lin1, lin2 are already on CUDA device.
This test did not actually do any CPU offloading. Adding it revealed an issue in no_shard (currently only when world_size == 1) case which is tracked for a fix in #74166 Differential Revision: [D35003793](https://our.internmc.facebook.com/intern/diff/D35003793/) [ghstack-poisoned]
facebook-github-bot
pushed a commit
that referenced
this pull request
Mar 24, 2022
Summary: Pull Request resolved: #74456 This test did not actually do any CPU offloading. Adding it revealed an issue in no_shard (currently only when world_size == 1) case which is tracked for a fix in #74166 ghstack-source-id: 152057449 Test Plan: CI Reviewed By: zhaojuanmao Differential Revision: D35003793 fbshipit-source-id: ffb5f8ea897cda84584ae83d362bea3c0c407c3b
Contributor
|
Hey @rohan-varma. |
shahofblah
pushed a commit
that referenced
this pull request
Mar 25, 2022
Summary: Pull Request resolved: #74456 This test did not actually do any CPU offloading. Adding it revealed an issue in no_shard (currently only when world_size == 1) case which is tracked for a fix in #74166 ghstack-source-id: 152057449 Test Plan: CI Reviewed By: zhaojuanmao Differential Revision: D35003793 fbshipit-source-id: ffb5f8ea897cda84584ae83d362bea3c0c407c3b (cherry picked from commit 06afeba)
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.
Stack from ghstack (oldest at bottom):
This test did not actually do any CPU offloading. Adding it revealed
an issue in no_shard (currently only when world_size == 1) case which is
tracked for a fix in #74166
Differential Revision: D35003793