[BUG or ENHANCEMENT] Update qk_layernorm. #210
+8
−2
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
Signed-off-by: ldwang <[email protected]>
| @@ -360,7 +360,7 @@ def __init__( | |||
| if submodules.q_layernorm is not None: | |||
| self.q_layernorm = build_module( | |||
| submodules.q_layernorm, | |||
| hidden_size=self.hidden_size_per_attention_head, | |||
| hidden_size=self.query_projection_size, | |||
There was a problem hiding this comment.
Thanks for your contribution. According to my understanding, the difference between the origin code and your code is as follows:
- origin code: normalize the every single head individually
- new code: normalize all the heads together
According to the input_layernormhttps://github.com/NVIDIA/Megatron-LM/blob/bbecd0812ffc9a90eef472fda91c60eead3f417f/megatron/core/transformer/transformer_layer.py#L111, we should normalize all the heads together. Maybe you can add an issue in Megatron-LM to recheck it.
I think you are right.
LGTM.
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Outdated suggestions cannot be applied.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.
With current qk_layernorm implement training did not converge.
One shared qk_layernorm acts on every head, however qk_layernorm should affect all heads.
So just enlarge the shape of qk_layernorm weights, training converges as expected.
List some models using qk_layernorm for references: