improve the eval perf with kv cache #3732
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/3732
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 6d92e53 with merge base 1343224 ( 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: D57764318 |
Summary:
The original implementation was too slow, and because of the frequent travel: cpu->gpu->cpu->gpu-> , it's inefficient. Change it to batch process the sequence so the compute remains in gpu
When evaluate stories model, before the change:
```
2024-05-23:23:42:25,115 INFO [evaluator.py:362] Running loglikelihood_rolling requests
100%|██████████| 5/5 [02:37<00:00, 31.50s/it]
wikitext: {'word_perplexity,none': 10589.525426446424, 'word_perplexity_stderr,none': 'N/A', 'byte_perplexity,none': 6.111053701258041, 'byte_perplexity_stderr,none': 'N/A', 'bits_per_byte,none': 2.6114211588515417, 'bits_per_byte_stderr,none': 'N/A', 'alias': 'wikitext'}
```
After the change:
```
2024-05-23:23:36:50,339 INFO [evaluator.py:362] Running loglikelihood_rolling requests
100%|██████████| 5/5 [00:03<00:00, 1.55it/s]
wikitext: {'word_perplexity,none': 10589.52618994558, 'word_perplexity_stderr,none': 'N/A', 'byte_perplexity,none': 6.111053787314264, 'byte_perplexity_stderr,none': 'N/A', 'bits_per_byte,none': 2.611421179167659, 'bits_per_byte_stderr,none': 'N/A', 'alias': 'wikitext'}
```
Differential Revision: D57764318
|
This pull request was exported from Phabricator. Differential Revision: D57764318 |
lucylq
approved these changes
May 24, 2024
|
This pull request has been merged in f42942a. |
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.
Summary:
The original implementation was too slow, and because of the frequent travel: cpu->gpu->cpu->gpu-> , it's inefficient. Change it to batch process the sequence so the compute remains in gpu
When evaluate stories model, before the change:
After the change:
Differential Revision: D57764318