MobileDiffusion: Rapid text-to-image generation on-device

Textual content-to-image diffusion models have proven distinctive capabilities in producing high-quality pictures from textual content prompts. Nevertheless, main fashions characteristic billions of parameters and are consequently costly to run, requiring highly effective desktops or servers (e.g., Stable Diffusion, DALL·E, and Imagen). Whereas current developments in inference options on Android through MediaPipe and iOS through Core ML have been made up to now 12 months, speedy (sub-second) text-to-image era on cell units has remained out of attain.

To that finish, in “MobileDiffusion: Subsecond Text-to-Image Generation on Mobile Devices”, we introduce a novel strategy with the potential for speedy text-to-image era on-device. MobileDiffusion is an environment friendly latent diffusion mannequin particularly designed for cell units. We additionally undertake DiffusionGAN to realize one-step sampling throughout inference, which fine-tunes a pre-trained diffusion mannequin whereas leveraging a GAN to mannequin the denoising step. Now we have examined MobileDiffusion on iOS and Android premium units, and it will probably run in half a second to generate a 512×512 high-quality picture. Its comparably small mannequin measurement of simply 520M parameters makes it uniquely suited to cell deployment.

Background

The relative inefficiency of text-to-image diffusion fashions arises from two major challenges. First, the inherent design of diffusion fashions requires iterative denoising to generate pictures, necessitating a number of evaluations of the mannequin. Second, the complexity of the community structure in text-to-image diffusion fashions includes a considerable variety of parameters, usually reaching into the billions and leading to computationally costly evaluations. Because of this, regardless of the potential advantages of deploying generative fashions on cell units, comparable to enhancing person expertise and addressing rising privateness issues, it stays comparatively unexplored throughout the present literature.

The optimization of inference effectivity in text-to-image diffusion fashions has been an energetic analysis space. Earlier research predominantly think about addressing the primary problem, looking for to scale back the variety of operate evaluations (NFEs). Leveraging superior numerical solvers (e.g., DPM) or distillation strategies (e.g., progressive distillation, consistency distillation), the variety of essential sampling steps have considerably diminished from a number of lots of to single digits. Some current strategies, like DiffusionGAN and Adversarial Diffusion Distillation, even scale back to a single essential step.

Nevertheless, on cell units, even a small variety of analysis steps might be gradual as a result of complexity of mannequin structure. To this point, the architectural effectivity of text-to-image diffusion fashions has obtained comparatively much less consideration. A handful of earlier works briefly touches upon this matter, involving the elimination of redundant neural community blocks (e.g., SnapFusion). Nevertheless, these efforts lack a complete evaluation of every element throughout the mannequin structure, thereby falling wanting offering a holistic information for designing extremely environment friendly architectures.

MobileDiffusion

Successfully overcoming the challenges imposed by the restricted computational energy of cell units requires an in-depth and holistic exploration of the mannequin’s architectural effectivity. In pursuit of this goal, our analysis undertakes an in depth examination of every constituent and computational operation inside Secure Diffusion’s UNet architecture. We current a complete information for crafting extremely environment friendly text-to-image diffusion fashions culminating within the MobileDiffusion.

The design of MobileDiffusion follows that of latent diffusion models. It incorporates three parts: a textual content encoder, a diffusion UNet, and a picture decoder. For the textual content encoder, we use CLIP-ViT/L14, which is a small mannequin (125M parameters) appropriate for cell. We then flip our focus to the diffusion UNet and picture decoder.

Diffusion UNet

As illustrated within the determine under, diffusion UNets generally interleave transformer blocks and convolution blocks. We conduct a complete investigation of those two basic constructing blocks. All through the examine, we management the coaching pipeline (e.g., knowledge, optimizer) to check the results of various architectures.

In traditional text-to-image diffusion fashions, a transformer block consists of a self-attention layer (SA) for modeling long-range dependencies amongst visible options, a cross-attention layer (CA) to seize interactions between textual content conditioning and visible options, and a feed-forward layer (FF) to post-process the output of consideration layers. These transformer blocks maintain a pivotal position in text-to-image diffusion fashions, serving as the first parts chargeable for textual content comprehension. Nevertheless, additionally they pose a big effectivity problem, given the computational expense of the eye operation, which is quadratic to the sequence size. We observe the thought of UViT structure, which locations extra transformer blocks on the bottleneck of the UNet. This design selection is motivated by the truth that the eye computation is much less resource-intensive on the bottleneck attributable to its decrease dimensionality.

Our UNet structure incorporates extra transformers within the center, and skips self-attention (SA) layers at larger resolutions.

Convolution blocks, particularly ResNet blocks, are deployed at every stage of the UNet. Whereas these blocks are instrumental for characteristic extraction and data move, the related computational prices, particularly at high-resolution ranges, might be substantial. One confirmed strategy on this context is separable convolution. We noticed that changing common convolution layers with light-weight separable convolution layers within the deeper segments of the UNet yields related efficiency.

Within the determine under, we evaluate the UNets of a number of diffusion fashions. Our MobileDiffusion displays superior effectivity by way of FLOPs (floating-point operations) and variety of parameters.

Comparability of some diffusion UNets.

Picture decoder

Along with the UNet, we additionally optimized the picture decoder. We educated a variational autoencoder (VAE) to encode an RGB picture to an 8-channel latent variable, with 8× smaller spatial measurement of the picture. A latent variable might be decoded to a picture and will get 8× bigger in measurement. To additional improve effectivity, we design a light-weight decoder structure by pruning the unique’s width and depth. The ensuing light-weight decoder results in a big efficiency increase, with almost 50% latency enchancment and higher high quality. For extra particulars, please discuss with our paper.

VAE reconstruction. Our VAE decoders have higher visible high quality than SD (Secure Diffusion).

One-step sampling

Along with optimizing the mannequin structure, we undertake a DiffusionGAN hybrid to realize one-step sampling. Coaching DiffusionGAN hybrid fashions for text-to-image era encounters a number of intricacies. Notably, the discriminator, a classifier distinguishing actual knowledge and generated knowledge, should make judgments based mostly on each texture and semantics. Furthermore, the price of coaching text-to-image fashions might be extraordinarily excessive, notably within the case of GAN-based fashions, the place the discriminator introduces extra parameters. Purely GAN-based text-to-image fashions (e.g., StyleGAN-T, GigaGAN) confront related complexities, leading to extremely intricate and costly coaching.

To beat these challenges, we use a pre-trained diffusion UNet to initialize the generator and discriminator. This design permits seamless initialization with the pre-trained diffusion mannequin. We postulate that the interior options throughout the diffusion mannequin comprise wealthy info of the intricate interaction between textual and visible knowledge. This initialization technique considerably streamlines the coaching.

The determine under illustrates the coaching process. After initialization, a loud picture is distributed to the generator for one-step diffusion. The result’s evaluated in opposition to floor fact with a reconstruction loss, just like diffusion mannequin coaching. We then add noise to the output and ship it to the discriminator, whose result’s evaluated with a GAN loss, successfully adopting the GAN to mannequin a denoising step. By utilizing pre-trained weights to initialize the generator and the discriminator, the coaching turns into a fine-tuning course of, which converges in lower than 10K iterations.

Illustration of DiffusionGAN fine-tuning.

Outcomes

Beneath we present instance pictures generated by our MobileDiffusion with DiffusionGAN one-step sampling. With such a compact mannequin (520M parameters in whole), MobileDiffusion can generate high-quality numerous pictures for numerous domains.

Photographs generated by our MobileDiffusion

We measured the efficiency of our MobileDiffusion on each iOS and Android units, utilizing totally different runtime optimizers. The latency numbers are reported under. We see that MobileDiffusion could be very environment friendly and may run inside half a second to generate a 512×512 picture. This lightning velocity probably permits many attention-grabbing use instances on cell units.

Latency measurements (s) on cell units.

Conclusion

With superior effectivity by way of latency and measurement, MobileDiffusion has the potential to be a really pleasant possibility for cell deployments given its functionality to allow a speedy picture era expertise whereas typing textual content prompts. And we’ll guarantee any utility of this expertise might be in-line with Google’s responsible AI practices.

Acknowledgments

We wish to thank our collaborators and contributors that helped deliver MobileDiffusion to on-device: Zhisheng Xiao, Yanwu Xu, Jiuqiang Tang, Haolin Jia, Lutz Justen, Daniel Fenner, Ronald Wotzlaw, Jianing Wei, Raman Sarokin, Juhyun Lee, Andrei Kulik, Chuo-Ling Chang, and Matthias Grundmann.