MEFT: Memory-Efficient Fine-Tuning through Sparse Adapter
This addresses the challenge of fine-tuning large models on knowledge-intensive tasks with limited GPU memory, though it is an incremental improvement over existing PEFT methods.
The paper tackles the problem of limited fine-tuning performance on complex tasks with Parameter-Efficient Fine-Tuning (PEFT) for Large Language Models by introducing a memory-efficient method that uses sparse adapters stored on CPU, achieving results comparable to larger memory setups under constrained resources like a 24GB GPU.
Parameter-Efficient Fine-tuning (PEFT) facilitates the fine-tuning of Large Language Models (LLMs) under limited resources. However, the fine-tuning performance with PEFT on complex, knowledge-intensive tasks is limited due to the constrained model capacity, which originates from the limited number of additional trainable parameters. To overcome this limitation, we introduce a novel mechanism that fine-tunes LLMs with adapters of larger size yet memory-efficient. This is achieved by leveraging the inherent activation sparsity in the Feed-Forward Networks (FFNs) of LLMs and utilizing the larger capacity of Central Processing Unit (CPU) memory compared to Graphics Processing Unit (GPU). We store and update the parameters of larger adapters on the CPU. Moreover, we employ a Mixture of Experts (MoE)-like architecture to mitigate unnecessary CPU computations and reduce the communication volume between the GPU and CPU. This is particularly beneficial over the limited bandwidth of PCI Express (PCIe). Our method can achieve fine-tuning results comparable to those obtained with larger memory capacities, even when operating under more limited resources such as a 24GB memory single GPU setup, with acceptable loss in training efficiency. Our codes are available at https://github.com/CURRENTF/MEFT.