Abstract: Systems and methods are provided for vector-quantized image modeling using vision transformers and improved codebook handling. In particular, the present disclosure provides a Vector-quantized Image Modeling (VIM) approach that involves pretraining a machine learning model (e.g., Transformer model) to predict rasterized image tokens autoregressively. The discrete image tokens can be encoded from a learned Vision-Transformer-based VQGAN (example implementations of which can be referred to as ViT-VQGAN). The present disclosure proposes multiple improvements over vanilla VQGAN from architecture to codebook learning, yielding better efficiency and reconstruction fidelity. The improved ViT-VQGAN further improves vector-quantized image modeling tasks, including unconditional image generation, conditioned image generation (e.g., class-conditioned image generation), and unsupervised representation learning.

Abstract: In a sampling circuit, a single cell in a series battery pack is isolated from a voltage divider resistor in a bleeder circuit by using a first isolation sampling switch, so as to prevent a drain current of the single cell. In addition, sampling errors in sampling voltages collected by the sampling circuit may be offset during differential calculation.

Abstract: A sampling circuit, an equalization circuit, and a system for a single cell in a series battery pack are provided. In the sampling circuit provided in this disclosure, the single cell is isolated from a voltage divider resistor in the bleeder circuit by using the first isolation sampling switch, so as to prevent a drain current of the single cell. In addition, sampling errors in sampling voltages collected by the sampling circuit may be offset during differential calculation.