[PGNCCL] Make NCCLComm extendable #142881
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]
pytorch-bot
bot
added
oncall: distributed
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/142881
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 109053e with merge base 5d3bc63 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR defines a set of APIs for interacting with `NCCLComm` (an existing abstraction of the real `ncclComm_t`).
This set of APIs include `create`, `split`, `destroy`, `dump`, etc, around the management of `NCCLComm`.
The previous implementation of the above methods under `NCCLComm` are now made the default impl of this API set.
And the impl of the API set is extendable out of trunk.
For example: for the implementation of dump
```
void dump(
std::shared_ptr<NCCLComm>& comm,
std::unordered_map<std::string, std::string>& dump) {
#if defined(IS_NCCLX) && defined(NCCL_COMM_DUMP)
if (comm->isAborted()) {
LOG(INFO) << "Communicator was aborted before trying to dump its state.";
return;
}
C10D_NCCL_CHECK(::ncclCommDump(ncclComm_, dump), std::nullopt);
#else
TORCH_CHECK(false, "NCCL version does not support dumping.");
#endif
}
```
The `#if defined(IS_NCCLX)` path can be now moved out of trunk.
cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o
[ghstack-poisoned]
Contributor
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
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Outdated suggestions cannot be applied.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 PR defines a set of APIs for interacting with
NCCLComm(an existing abstraction of the realncclComm_t).This set of APIs include
create,split,destroy,dump, etc, around the management ofNCCLComm.The previous implementation of the above methods under
NCCLCommare now made the default impl of this API set.And the impl of the API set is extendable out of trunk.
For example: for the implementation of dump
One can now host the
#if defined(IS_NCCLX)path out of trunk.TODO:
ncclConfig_tstill shows up in API surface. We probably need a way to make it opaque.cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @c-p-i-o