TensorX
返回文献探索

Paper · arXiv 2406.06525

Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation

Peize Sun, Yi Jiang, Shoufa Chen, Shilong Zhang, Bingyue Peng, Ping Luo, Zehuan Yuan

71 upvotesJune 10, 2024arXiv 预印本
AI 摘要

LlamaGen applies the next-token prediction paradigm from large language models to image generation, achieving state-of-the-art performance with various model sizes and conditions.

next-token predictionlarge language modelsautoregressive modelsimage tokenizersclass-conditional image generationtext-conditional image generationFIDrFIDdiffusion modelsLDMDiTmultimodal foundation models

Abstract

We introduce LlamaGen, a new family of image generation models that apply original ``next-token prediction'' paradigm of large language models to visual generation domain. It is an affirmative answer to whether vanilla autoregressive models, e.g., Llama, without inductive biases on visual signals can achieve state-of-the-art image generation performance if scaling properly. We reexamine design spaces of image tokenizers, scalability properties of image generation models, and their training data quality. The outcome of this exploration consists of: (1) An image tokenizer with downsample ratio of 16, reconstruction quality of 0.94 rFID and codebook usage of 97% on ImageNet benchmark. (2) A series of class-conditional image generation models ranging from 111M to 3.1B parameters, achieving 2.18 FID on ImageNet 256x256 benchmarks, outperforming the popular diffusion models such as LDM, DiT. (3) A text-conditional image generation model with 775M parameters, from two-stage training on LAION-COCO and high aesthetics quality images, demonstrating competitive performance of visual quality and text alignment. (4) We verify the effectiveness of LLM serving frameworks in optimizing the inference speed of image generation models and achieve 326% - 414% speedup. We release all models and codes to facilitate open-source community of visual generation and multimodal foundation models.

北京市昌平区探索星信息技术及软件开发工作室

京ICP备2026059466号