EMLoC: Emulator-based Memory-efficient Fine-tuning with LoRA Correction

TL;DR

EMLoC introduces a memory-efficient fine-tuning method for large models using emulator construction and LoRA correction, enabling cost-effective adaptation on standard hardware.

Contribution

The paper presents a novel emulator-based fine-tuning framework with a compensation algorithm, reducing memory requirements and allowing large model fine-tuning on consumer GPUs.

Findings

Outperforms baseline methods across multiple datasets and modalities.

Enables fine-tuning of a 38B model on a 24GB GPU without quantization.

Supports flexible compression ratios and standard training pipelines.

Abstract

Open-source foundation models have seen rapid adoption and development, enabling powerful general-purpose capabilities across diverse domains. However, fine-tuning large foundation models for domain-specific or personalized tasks remains prohibitively expensive for most users due to the significant memory overhead beyond that of inference. We introduce EMLoC, an Emulator-based Memory-efficient fine-tuning framework with LoRA Correction, which enables model fine-tuning within the same memory budget required for inference. EMLoC constructs a task-specific light-weight emulator using activation-aware singular value decomposition (SVD) on a small downstream calibration set. Fine-tuning then is performed on this lightweight emulator via LoRA. To tackle the misalignment between the original model and the compressed emulator, we propose a novel compensation algorithm to correct the fine-tuned LoRA module, which thus can be merged into the original model for inference. EMLoC supports flexible compression ratios and standard training pipelines, making it adaptable to a wide range of applications. Extensive experiments demonstrate that EMLoC outperforms other baselines across multiple datasets and modalities. Moreover, without quantization, EMLoC enables fine-tuning of a 38B model, which originally required 95GB of memory, on a single 24GB consumer GPU-bringing efficient and practical model adaptation to individual users.