Closed
Conversation
This comment was marked as spam.
This comment was marked as spam.
Sorry, something went wrong.
This comment was marked as spam.
This comment was marked as spam.
Sorry, something went wrong.
zhaochenyang20
requested changes
Jun 8, 2025
fzyzcjy
reviewed
Jun 8, 2025
Collaborator
fzyzcjy
left a comment
There was a problem hiding this comment.
not yet fully reviewed, just spent a few minute to glance
fzyzcjy
reviewed
Jun 9, 2025
| self.tp_worker.worker.model_runner.model | ||
| ) | ||
|
|
||
| self.weights_memory_saver_adapter.check_validity( |
Collaborator
There was a problem hiding this comment.
nit: to be honest I feel the PR to be a little bit over complicated, if time permits, could you please update them a bit
Collaborator
Author
There was a problem hiding this comment.
simplified, let me know if its better now
10 tasks
10 tasks
Collaborator
Author
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
To Reviewer: Need this PR to check in first: fzyzcjy/torch_memory_saver#17
Currently, many unit tests are failed because tms only support singleton without above change.
Related PR:
Thanks a lot for @fzyzcjy 's guidance and help.
Motivation
To close: fzyzcjy/torch_memory_saver#3
vLLM has supported Multi-Stage Awake for vllm engine: vllm-project/vllm#15254
But in SGLang, we are using
torch_memory_saverfor holding the model weight and KV Cache virtual address (to make sure CUDA Graph works across different rollouts)And
torch_memory_saveris a singleton, which make it hard for SGLang to support Multi-Stage Awake which is critical in RL use caseWithout this feature, we can only set KV Cache mem frac rate to <= 0.7 or even lower (e.g 0.3)
Need this PR to check in first: fzyzcjy/torch_memory_saver#17
Modifications
Checklist