Added Functionality for LoRA Finetuning #218
Open
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
It is the base file which implements the LoRA Linear Layer and implements useful methods for working with a model on which LoRA has been applied. Furthermore, implements a function which can convert any model into one with LoRA if the names of the affected linear layers of the model are known
Added arguments for a flag for applying lora and added functionality to change the parameters used in lora like rank, alpha, dropout probability, target module names (the module names which will be changed to lora linear from linear) and lora weights path (the path to the lora weights)
Changed the training script to allow LoRA implementation. Three main areas of changes * Imported apply_lora_to_model * Applied the above function to the model if the user_lora flag is True * Saved the LoRA weights at the checkpoints when the other model weights are being saved
Edited training script so that LoRA can be applied to the model
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Outdated suggestions cannot be applied.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.
Allows functionality for finetuning the model using LoRA. To make use of the changes, need to only add extrac flags while running the training script. The flags would be the following:
-- use_lora True --lora_rank 8 --lora_alpha 16 --lora_dropout 0.05 --lora_target_modules q_proj k_proj v_proj out_proj fc1 fc2 --lora_weights_path