Rename linear projection #37
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
… static method to create a Dense layer from ST weights
raphaelsty
reviewed
Aug 20, 2024
| model = Dense(**config) | ||
| if os.path.exists(os.path.join(input_path, "model.safetensors")): | ||
| load_safetensors_model(model, os.path.join(input_path, "model.safetensors")) | ||
| else: |
There was a problem hiding this comment.
You can remove the else and return the model in the if after load_safetensors_model
It will enable to designed model_load_state_dict
raphaelsty
reviewed
Aug 20, 2024
| @@ -294,6 +316,18 @@ def __init__( | |||
| def load(input_path) -> "ColBERT": | |||
| return ColBERT(model_name_or_path=input_path) | |||
|
|
|||
| def num_modules(self) -> int: | |||
There was a problem hiding this comment.
You can rename this function by
def __len__(self) -> int:
return len(_self.modules)then you can call len(self) to get the actual number of modules
raphaelsty
reviewed
Aug 20, 2024
| and self[1].get_sentence_embedding_dimension() != embedding_size | ||
| ): | ||
| logger.warning( | ||
| f"The checkpoint contains a dense layer but with incorrect dimension. Replacing it with a Dense layer with output dimensions ({hidden_size}, {embedding_size})" |
There was a problem hiding this comment.
It would be great to print the actual embedding dimension you found self[1].get_sentence_embedding_dimension()
raphaelsty
reviewed
Aug 20, 2024
| def num_modules(self) -> int: | ||
| return len(self._modules) | ||
|
|
||
| def convert_dense_layer_from_sentence_transformer( |
There was a problem hiding this comment.
Could you add a docstring here ?
raphaelsty
reviewed
Aug 20, 2024
| return features | ||
|
|
||
| @staticmethod | ||
| def from_sentence_transformers(dense_st: DenseSentenceTransformer): |
There was a problem hiding this comment.
Might be worth to add a docstring here to explain that the Sentence Transformer has an activation function and we don't want one
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.
This PR refactors the old LinearProjection module to better extend the existing sentence transformer layer.
This allows to unify the names so the conversion from one lib checkpoint to another is easier (as the only difference here is in the activation function and the forward function being called on all the tokens instead of just the pooled one).
I changed a bit the behavior of the model creation: