Add ARA and ARA LoRA abliteration support for HF and EXL3 backends#35
Merged
Conversation
Implement Arbitrary-Rank Ablation (ARA) and its LoRA variant as alternative abliteration methods alongside the existing directional approach. ARA LoRA optimises low-rank adapter factors directly rather than modifying full weight matrices, making it compatible with quantised models (bnb 4-bit, EXL3) and eliminating the need for full model reloads between trials. ARA LoRA implementation based on work by kabachuha (p-e-w#332). New config options: use_ara, use_ara_lora, ara_lora_rank. https://claude.ai/code/session_01LhXpnSi1t4CzjY2DQEeA9k
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Implement Arbitrary-Rank Ablation (ARA) and its LoRA variant as alternative abliteration methods alongside the existing directional approach. ARA LoRA optimises low-rank adapter factors directly rather than modifying full weight matrices, making it compatible with quantized models (bnb 4-bit, EXL3) and eliminating the need for full model reloads between trials.
ARA LoRA implementation based on work by kabachuha (p-e-w#332).
New config options: use_ara, use_ara_lora, ara_lora_rank.