Diffusion models create data from noise by inverting the forward paths of data towards noise and have emerged as a powerful generative modeling technique for high-dimensional, perceptual data such as images and videos. Rectified flow is a recent generative model formulation that connects data and noise in a straight line. Despite its better theoretical properties and conceptual simplicity, it is not yet decisively established as standard practice. In this work, we improve existing noise sampling techniques for training rectified flow models by biasing them towards perceptually relevant scales. Through a large-scale study, we demonstrate the superior performance of this approach compared to established diffusion formulations for high-resolution text-to-image synthesis. Additionally, we present a novel transformer-based architecture for text-to-image generation that uses separate weights for the two modalities and enables a bidirectional flow of information between image and text tokens, improving text comprehension, typography, and human preference ratings. We demonstrate that this architecture follows predictable scaling trends and correlates lower validation loss to improved text-to-image synthesis as measured by various metrics and human evaluations. Our largest models outperform state-of-the-art models. Stability AI is considering making experimental data, code, and model weights publicly available.

By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs.

Large intra-class variation is the result of changes in multiple object characteristics. Images, however, only show the superposition of different variable factors such as appearance or shape. Therefore, learning to disentangle and represent these different characteristics poses a great challenge, especially in the unsupervised case. Moreover, large object articulation calls for a flexible part-based model. We present an unsupervised approach for disentangling appearance and shape by learning parts consistently over all instances of a category. Our model for learning an object representation is trained by simultaneously exploiting invariance and equivariance constraints between synthetically transformed images. Since no part annotation or prior information on an object class is required, the approach is applicable to arbitrary classes. We evaluate our approach on a wide range of object categories and diverse tasks including pose prediction, disentangled image synthesis, and video-to-video translation. The approach outperforms the state-of-the-art on unsupervised keypoint prediction and compares favorably even against supervised approaches on the task of shape and appearance transfer.

Since the introduction of lead-free solder alloys, the impact of voids in solder joints has rapidly gained in importance. Particularly in power electronics, voids in solder joints can reduce their reliability due to less heat dissipation and a significant reduction of the cross-sectional area. While void formation in the printed circuit board (PCB) technology has been the focus of many research projects, the increasing application of molded interconnect devices (MIDs) demands a vivid understanding of void formation and its impact on reliability. This report will show that those processes specific to MID technology do not significantly influence the void formation in SAC305 solder deposits. The processed different roughness levels on Vectra, Vestamid, and FR4 substrate materials show no significant impact on void formation. The surface finish metallization was detected as the most important factor. Especially, the chemical-Sn (iSn) surface finish is inclined to having more voids compared to a nickel-phosphor-gold (NiP-Au) finish. Furthermore, three solder processes were investigated: convection, condensation, and vacuum-condensation soldering. The vacuum-condensation processes significantly reduced the voids in the Sn-3.0Ag-0.5Cu (simplified as SAC305) solder joints of CR0805, MLF20, and switches on all the investigated substrates. An accelerated thermal cycling test of these components according to JESD22-A104D demonstrates no significant effect of the void content of <;19% on the solder joints' reliability. Finally, the result of accelerated thermal cycling shows the possibility to achieve an equivalent lifetime of the solder joints on Vectra in comparison to the FR4 substrate.

Aging, which can be regarded as a time-dependent variability, has until recently not received much attention in the field of electronic design automation. This is changing because increasing reliability costs threaten the continued scaling of ICs. We investigate the impact of aging effects on single combinatorial gates and present methods that help to reduce the reliability costs by accurately analyzing the performance degradation of aged circuits at gate and macro cell level.

We present an aging analysis flow able to calculate the degraded circuit timing. To the best of our knowledge it is the first approach on gate level so far capable of analyzing the impact of the two dominant drift-related aging effects - NBTI and HCI - on complex digital circuits. The aging-aware gate model used to compute the aged circuit timing provides not just the cell delay degradation, but also the degradation of the output slope. To get more accurate results, the individual workload of a gate can be considered.