Jeesup/svd-safety-l2_swift_jbbsft1_remove50
Jeesup/svd-safety-l2_swift_jbbsft1_remove50 is a 7 billion parameter Llama-2-7b-chat checkpoint compressed with Swift-SVD to 50% of its original parameters and then recovered with SVD-LLM's stage-2 LoRA. This model is a research artifact designed to study how SVD compression impacts safety behavior and the effectiveness of component-selection rules for recovery. It is specifically intended for evaluating safety/utility trade-offs under compression, not as a general-purpose chat model.
Loading preview...
Model Overview
Jeesup/svd-safety-l2_swift_jbbsft1_remove50 is a research artifact derived from the meta-llama/Llama-2-7b-chat-hf base model. It has undergone significant compression using Swift-SVD, reducing its parameters to 50% of the original 7 billion, followed by recovery using SVD-LLM's stage-2 LoRA. This specific configuration uses dynamic rank allocation with an alpha of 0.6 and WikiText2 calibration.
Key Characteristics
- Compression Method: Swift-SVD (dynamic rank allocation, alpha 0.6, 256 x 2048 WikiText2 calibration).
- Parameter Reduction: Compressed to 50.0% of the original dense parameters.
- Recovery Method: SVD-LLM's stage-2 LoRA (sequential U then V, r=8, alpha=16, 2 epochs per half, lr 0.0001, batch 64, cutoff 256).
- Context Length: 4096 tokens.
Intended Use and Limitations
This model is not a general-purpose chat model. It is a specific experimental subject from a study investigating the impact of SVD compression on safety behavior and the efficacy of various recovery techniques. The model's safety behavior is deliberately degraded relative to the original Llama-2-7b-chat due to compression. Users should treat this as an experimental subject for research purposes, particularly for quantifying safety/utility trade-offs under compression, and not as a deployable assistant. Evaluation of its properties is crucial before drawing conclusions.