M6: A Chinese Multimodal Pretrainer

TL;DR

This paper introduces M6, the largest Chinese multimodal pretraining model with over 10 billion parameters, demonstrating superior performance across various tasks including text-guided image generation.

Contribution

The work presents the largest Chinese multimodal dataset and a novel unified pretraining method, scaling the model to 100 billion parameters for enhanced performance.

Findings

Outperforms strong baselines in downstream tasks

Achieves high-quality, high-resolution text-guided image generation

Demonstrates the effectiveness of large-scale multimodal pretraining in Chinese

Abstract

In this work, we construct the largest dataset for multimodal pretraining in Chinese, which consists of over 1.9TB images and 292GB texts that cover a wide range of domains. We propose a cross-modal pretraining method called M6, referring to Multi-Modality to Multi-Modality Multitask Mega-transformer, for unified pretraining on the data of single modality and multiple modalities. We scale the model size up to 10 billion and 100 billion parameters, and build the largest pretrained model in Chinese. We apply the model to a series of downstream applications, and demonstrate its outstanding performance in comparison with strong baselines. Furthermore, we specifically design a downstream task of text-guided image generation, and show that the finetuned M6 can create high-quality images with high resolution and abundant details.