[ET][Memory planning] Improve greedy memory planning. #7926
Conversation
This file contains 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 diff replaces the old greedy algorithm. Older algorithm resulted in 35% worse compared to theoretical optimum. THis matter for long context even more since additional overhead can be few hundred MB. For example the theorical optimial for llama3_2 8B, 4-bit quantized modelw ith context length of 2k needs about 1G of memory. This theoretcial max can be observed by looking at the peaks in memory profile. Current agorithm resulted in about 1.6GB of planned memory. New algorithm reduce that to about 1.1G. Differential Revision: [D68448332](https://our.internmc.facebook.com/intern/diff/D68448332/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7926
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit 290ee8d with merge base d4a8f8f ( BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D68448332 |
This was referenced Jan 24, 2025
This diff replaces the old greedy algorithm. Older algorithm resulted in 35% worse compared to theoretical optimum. THis matter for long context even more since additional overhead can be few hundred MB. For example the theorical optimial for llama3_2 8B, 4-bit quantized modelw ith context length of 2k needs about 1G of memory. This theoretcial max can be observed by looking at the peaks in memory profile. Current agorithm resulted in about 1.6GB of planned memory. New algorithm reduce that to about 1.1G. Differential Revision: [D68448332](https://our.internmc.facebook.com/intern/diff/D68448332/) [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D68448332 |
This diff replaces the old greedy algorithm. Older algorithm resulted in 35% worse compared to theoretical optimum. THis matter for long context even more since additional overhead can be few hundred MB. For example the theorical optimial for llama3_2 8B, 4-bit quantized modelw ith context length of 2k needs about 1G of memory. This theoretcial max can be observed by looking at the peaks in memory profile. Current agorithm resulted in about 1.6GB of planned memory. New algorithm reduce that to about 1.1G. Differential Revision: [D68448332](https://our.internmc.facebook.com/intern/diff/D68448332/) [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D68448332 |
kimishpatel
added
module: exir
This diff replaces the old greedy algorithm. Older algorithm resulted in 35% worse compared to theoretical optimum. THis matter for long context even more since additional overhead can be few hundred MB. For example the theorical optimial for llama3_2 8B, 4-bit quantized modelw ith context length of 2k needs about 1G of memory. This theoretcial max can be observed by looking at the peaks in memory profile. Current agorithm resulted in about 1.6GB of planned memory. New algorithm reduce that to about 1.1G. Differential Revision: [D68448332](https://our.internmc.facebook.com/intern/diff/D68448332/) cc JacobSzwejbka angelayi [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D68448332 |
This diff replaces the old greedy algorithm. Older algorithm resulted in 35% worse compared to theoretical optimum. THis matter for long context even more since additional overhead can be few hundred MB. For example the theorical optimial for llama3_2 8B, 4-bit quantized modelw ith context length of 2k needs about 1G of memory. This theoretcial max can be observed by looking at the peaks in memory profile. Current agorithm resulted in about 1.6GB of planned memory. New algorithm reduce that to about 1.1G. Differential Revision: [D68448332](https://our.internmc.facebook.com/intern/diff/D68448332/) cc JacobSzwejbka angelayi [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D68448332 |
GregoryComer
approved these changes
Jan 28, 2025
kimishpatel
temporarily deployed
to
upload-benchmark-results
GitHub Actions
Inactive
facebook-github-bot
merged commit 8d4749e
into
gh/kimishpatel/151/base
50 of 53 checks passed
facebook-github-bot
temporarily deployed
to
cherry-pick-bot
GitHub Actions
Inactive
manuelcandales
pushed a commit
that referenced
this pull request
Jan 28, 2025
* Fix memory profiling for memory.view ops Pull Request resolved: #7925 ATT ghstack-source-id: 263342054 @exported-using-ghexport Differential Revision: [D68448333](https://our.internmc.facebook.com/intern/diff/D68448333/) * [ET][Memory planning] Improve greedy memory planning. Pull Request resolved: #7926 This diff replaces the old greedy algorithm. Older algorithm resulted in 35% worse compared to theoretical optimum. THis matter for long context even more since additional overhead can be few hundred MB. For example the theorical optimial for llama3_2 8B, 4-bit quantized modelw ith context length of 2k needs about 1G of memory. This theoretcial max can be observed by looking at the peaks in memory profile. Current agorithm resulted in about 1.6GB of planned memory. New algorithm reduce that to about 1.1G. ghstack-source-id: 263342052 @exported-using-ghexport Differential Revision: [D68448332](https://our.internmc.facebook.com/intern/diff/D68448332/) --------- Co-authored-by: Kimish Patel <[email protected]>
YIWENX14
pushed a commit
that referenced
this pull request
Jan 28, 2025
* Fix memory profiling for memory.view ops Pull Request resolved: #7925 ATT ghstack-source-id: 263342054 @exported-using-ghexport Differential Revision: [D68448333](https://our.internmc.facebook.com/intern/diff/D68448333/) * [ET][Memory planning] Improve greedy memory planning. Pull Request resolved: #7926 This diff replaces the old greedy algorithm. Older algorithm resulted in 35% worse compared to theoretical optimum. THis matter for long context even more since additional overhead can be few hundred MB. For example the theorical optimial for llama3_2 8B, 4-bit quantized modelw ith context length of 2k needs about 1G of memory. This theoretcial max can be observed by looking at the peaks in memory profile. Current agorithm resulted in about 1.6GB of planned memory. New algorithm reduce that to about 1.1G. ghstack-source-id: 263342052 @exported-using-ghexport Differential Revision: [D68448332](https://our.internmc.facebook.com/intern/diff/D68448332/) --------- Co-authored-by: Kimish Patel <[email protected]>
zonglinpeng
pushed a commit
to zonglinpeng/executorch
that referenced
this pull request
Jan 30, 2025
* Fix memory profiling for memory.view ops Pull Request resolved: pytorch#7925 ATT ghstack-source-id: 263342054 @exported-using-ghexport Differential Revision: [D68448333](https://our.internmc.facebook.com/intern/diff/D68448333/) * [ET][Memory planning] Improve greedy memory planning. Pull Request resolved: pytorch#7926 This diff replaces the old greedy algorithm. Older algorithm resulted in 35% worse compared to theoretical optimum. THis matter for long context even more since additional overhead can be few hundred MB. For example the theorical optimial for llama3_2 8B, 4-bit quantized modelw ith context length of 2k needs about 1G of memory. This theoretcial max can be observed by looking at the peaks in memory profile. Current agorithm resulted in about 1.6GB of planned memory. New algorithm reduce that to about 1.1G. ghstack-source-id: 263342052 @exported-using-ghexport Differential Revision: [D68448332](https://our.internmc.facebook.com/intern/diff/D68448332/) --------- Co-authored-by: Kimish Patel <[email protected]>
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 diff replaces the old greedy algorithm. Older algorithm resulted in 35%
worse compared to theoretical optimum. THis matter for long context even more
since additional overhead can be few hundred MB.
For example the theorical optimial for llama3_2 8B, 4-bit quantized modelw ith
context length of 2k needs about 1G of memory. This theoretcial max can be
observed by looking at the peaks in memory profile.
Current agorithm resulted in about 1.6GB of planned memory. New algorithm
reduce that to about 1.1G.
Differential Revision: D68448332
cc @JacobSzwejbka @angelayi