[skip ci][ci] Reduce GPU shards#11463
Closed
driazati wants to merge 1 commit intoapache:mainfrom
driazati:less_gpu
Closed
[skip ci][ci] Reduce GPU shards#11463driazati wants to merge 1 commit intoapache:mainfrom driazati:less_gpu
driazati wants to merge 1 commit intoapache:mainfrom
driazati:less_gpu
Conversation
More shard rebalancing, while using more GPU shards in experiments reduced CI runtime, in prod these ended up eating all the available GPU capacity and causing queuing delays such that overall runtime was basically the same, but with the added cost of the per-node setup * the extra nodes.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
More shard rebalancing, while using more GPU shards in experiments reduced CI runtime, in prod these ended up eating all the available GPU capacity and causing queuing delays such that overall runtime was basically the same, but with the added cost of the per-node setup * the extra nodes.
Thanks for contributing to TVM! Please refer to guideline https://tvm.apache.org/docs/contribute/ for useful information and tips. After the pull request is submitted, please request code reviews from Reviewers by @ them in the pull request thread.