This March 2023 paper introduces PETALS, a novel system designed to facilitate the collaborative inference and fine-tuning of large language models (LLMs) by pooling resources from multiple participants. It addresses the significant computational and memory demands of LLMs, which typically restrict access for many researchers. PETALS proposes an alternative to traditional methods like slow RAM offloading or inflexible inference APIs by allowing distributed processing across a network of consumer GPUs, enhancing speed and flexibility. The system incorporates optimizations like 8-bit quantization and dynamic load balancing to improve performance and reliability. Ultimately, PETALS aims to democratize access to powerful LLMs, enabling broader research and application development that was previously cost-prohibitive.
Source:
https://arxiv.org/pdf/2209.01188
