support GQA #7906
Merged
support GQA #7906
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
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## develop #7906 +/- ##
========================================
Coverage 56.56% 56.56%
========================================
Files 589 589
Lines 89939 89964 +25
========================================
+ Hits 50874 50891 +17
- Misses 39065 39073 +8 ☔ View full report in Codecov by Sentry. |
ZHUI
reviewed
Feb 21, 2024
Comment on lines
+936
to
+953
| query_states, _, _ = fused_rotary_position_embedding( | ||
| query_states, | ||
| None, | ||
| None, | ||
| sin=sin, | ||
| cos=cos, | ||
| position_ids=position_ids, | ||
| use_neox_rotary_style=False, | ||
| ) | ||
| key_states, _, _ = fused_rotary_position_embedding( | ||
| key_states, | ||
| None, | ||
| None, | ||
| sin=sin, | ||
| cos=cos, | ||
| position_ids=position_ids, | ||
| use_neox_rotary_style=False, | ||
| ) |
There was a problem hiding this comment.
GQA的时候，原来的代码 用 fused_rotary_position_embedding 是有问题的吗？
There was a problem hiding this comment.
是的，Paddle旧版本的fused_rotary_position_embedding不支持传入的q和k/v 有不同的heads，所以等效的方式是单独处理q，k，需要分别调用2次接口。
我们在dev已经做了支持，所以可以直接调用1次接口。
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Outdated suggestions cannot be applied.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.
PR types
Performance optimizationPR changes
OthersDescription
support GQA目前Paddle的develop已经为fa和fuse_rope支持gqa/mqa，但2.6及以下版本未支持，因此为模型实现兼容性修改。
设置num_key_value_heads = 8，使用Paddle2.6版本和Paddle-dev分别测试llama2-13b模型收敛情况。
