SnapGen: Taming High-Resolution Text-to-Image Models for Mobile Devices with Efficient Architectures and Training
This addresses the challenge of deploying high-resolution text-to-image models on resource-constrained mobile devices, representing an incremental improvement in efficiency and speed.
The paper tackled the problem of large, slow, and low-quality text-to-image generation on mobile devices by developing SnapGen, an efficient model that generates 1024x1024 px images in about 1.4 seconds on mobile, achieving an FID of 2.06 on ImageNet-1K with only 372M parameters and surpassing larger models on benchmarks.
Existing text-to-image (T2I) diffusion models face several limitations, including large model sizes, slow runtime, and low-quality generation on mobile devices. This paper aims to address all of these challenges by developing an extremely small and fast T2I model that generates high-resolution and high-quality images on mobile platforms. We propose several techniques to achieve this goal. First, we systematically examine the design choices of the network architecture to reduce model parameters and latency, while ensuring high-quality generation. Second, to further improve generation quality, we employ cross-architecture knowledge distillation from a much larger model, using a multi-level approach to guide the training of our model from scratch. Third, we enable a few-step generation by integrating adversarial guidance with knowledge distillation. For the first time, our model SnapGen, demonstrates the generation of 1024x1024 px images on a mobile device around 1.4 seconds. On ImageNet-1K, our model, with only 372M parameters, achieves an FID of 2.06 for 256x256 px generation. On T2I benchmarks (i.e., GenEval and DPG-Bench), our model with merely 379M parameters, surpasses large-scale models with billions of parameters at a significantly smaller size (e.g., 7x smaller than SDXL, 14x smaller than IF-XL).