Advancing Model Refinement: Muon-Optimized Distillation and Quantization for LLM Deployment
This addresses resource constraints for deploying LLMs on edge devices, but it is incremental as it builds on existing techniques like GPTQ and LoRA.
The paper tackles deployment challenges of large language models on edge devices by proposing an integrated framework combining quantization, low-rank adaptation, and data distillation, achieving up to 2x memory compression and improved performance on benchmarks compared to baseline methods.
Large Language Models (LLMs) enable advanced natural language processing but face deployment challenges on resource-constrained edge devices due to high computational, memory, and energy demands. Optimizing these models requires addressing three key challenges: acquiring task-specific data, fine-tuning for performance, and compressing models to accelerate inference while reducing resource demands. We propose an integrated framework combining GPTQ-based quantization, low-rank adaptation (LoRA), and a specialized data distillation process to significantly reduce model size and complexity while preserving or enhancing task-specific performance. By leveraging data distillation, knowledge distillation via Kullback-Leibler divergence, Bayesian hyperparameter optimization, and the Muon optimizer, our pipeline achieves up to 2x memory compression (e.g., reducing a 6GB model to 3GB) and enables efficient inference for specialized tasks. Empirical results demonstrate superior performance on standard LLM benchmarks compared to GPTQ quantization alone, with the Muon optimizer notably enhancing fine-tuned models' resistance to accuracy decay during quantization.