Fix T5 adapter layer input #479
Merged
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
Clarify adapter layer input parameters.
hSterz
reviewed
Jan 17, 2023
| @@ -340,7 +340,7 @@ def __init__(self, config: T5Config): | |||
| def forward(self, hidden_states): | |||
| forwarded_states = self.layer_norm(hidden_states) | |||
| forwarded_states = self.DenseReluDense(forwarded_states) | |||
| hidden_states = self.adapter_layer_forward(hidden_states, self.dropout(forwarded_states), None) | |||
| hidden_states = self.adapter_layer_forward(self.dropout(forwarded_states), hidden_states, None) | |||
There was a problem hiding this comment.
We could make it super obvious which one the residual connection is by using hidden_states=..., residual_connection=... in the method call. This could help people adding new models when they have a look at the currently implemented models.
hSterz
approved these changes
Jan 17, 2023
This pull request was closed.
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Outdated suggestions cannot be applied.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.
The adapter implementation of T5 currently passes the inputs to the adapter layer in the wrong order (i.e. residual inputs before hidden states). Thus, the sequential adapter implementation is effectively a parallel adapter.
This PR fixes the inputs order. It also clarifies the adapter layer input parameters.