[Core][Optimization] remove vllm-nccl #5091
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
simon-mo
approved these changes
May 29, 2024
chengzhi-lu
pushed a commit
to chengzhi-lu/vllm
that referenced
this pull request
May 29, 2024
dtrifiro
pushed a commit
to opendatahub-io/vllm
that referenced
this pull request
May 31, 2024
Closed
robertgshaw2-redhat
pushed a commit
to neuralmagic/nm-vllm
that referenced
this pull request
Jun 8, 2024
dtrifiro
added a commit
to dtrifiro/vllm
that referenced
this pull request
Jun 13, 2024
Fixed upstream in vllm-project#5091
joerunde
pushed a commit
to joerunde/vllm
that referenced
this pull request
Jun 17, 2024
dtrifiro
added a commit
to opendatahub-io/vllm
that referenced
this pull request
Jun 18, 2024
Fixed upstream in vllm-project#5091
dtrifiro
added a commit
to opendatahub-io/vllm
that referenced
this pull request
Jun 21, 2024
Fixed upstream in vllm-project#5091
prashantgupta24
pushed a commit
to opendatahub-io/vllm
that referenced
this pull request
Jul 1, 2024
Fixed upstream in vllm-project#5091
robertgshaw2-redhat
pushed a commit
to neuralmagic/nm-vllm
that referenced
this pull request
Jul 14, 2024
dtrifiro
added a commit
to opendatahub-io/vllm
that referenced
this pull request
Jul 17, 2024
Fixed upstream in vllm-project#5091
dtrifiro
added a commit
to opendatahub-io/vllm
that referenced
this pull request
Jul 23, 2024
Fixed upstream in vllm-project#5091
dtrifiro
added a commit
to opendatahub-io/vllm
that referenced
this pull request
Aug 6, 2024
Fixed upstream in vllm-project#5091
dtrifiro
added a commit
to dtrifiro/vllm
that referenced
this pull request
Sep 13, 2024
Fixed upstream in vllm-project#5091
dtrifiro
added a commit
to opendatahub-io/vllm
that referenced
this pull request
Sep 13, 2024
Fixed upstream in vllm-project#5091
dtrifiro
added a commit
to opendatahub-io/vllm
that referenced
this pull request
Sep 27, 2024
Fixed upstream in vllm-project#5091
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
After months of investigation, I finally find the root cause is that NCCL 2.19 will turn on virtual memory by default, which costs memory during cudagraph capture.
Per the documentation:
The transition happens right in nccl 2.19, where this feature becomes on by default. And vLLM, very unfortunately, hits exactly the problem when we upgrade to pytorch 2.2.
The solution is simple, just adding a new environment variable, and it's fine. We can use the nccl library brought by pytorch. And we don't need to ship
vllm-ncclanymore.