Jun Zhang

Jun Zhang, Jue Wang, Huan Li et al.

We present a novel inference scheme, self-speculative decoding, for accelerating Large Language Models (LLMs) without the need for an auxiliary model. This approach is characterized by a two-stage process: drafting and verification. The drafting stage generates draft tokens at a slightly lower quality but more quickly, which is achieved by selectively skipping certain intermediate layers during drafting. Subsequently, the verification stage employs the original LLM to validate those draft output tokens in one forward pass. This process ensures the final output remains identical to that produced by the unaltered LLM. Moreover, the proposed method requires no additional neural network training and no extra memory footprint, making it a plug-and-play and cost-effective solution for inference acceleration. Benchmarks with LLaMA-2 and its variants demonstrated a speedup up to 1.99$\times$.

Xiaorui Bai, Wenyong Wang, Jun Zhang et al.

Flow field segmentation and classification help researchers to understand vortex structure and thus turbulent flow. Existing deep learning methods mainly based on global information and focused on 2D circumstance. Based on flow field theory, we propose novel flow field segmentation and classification deep learning methods in three-dimensional space. We construct segmentation criterion based on local velocity information and classification criterion based on the relationship between local vorticity and vortex wake, to identify vortex structure in 3D flow field, and further classify the type of vortex wakes accurately and rapidly. Simulation experiment results showed that, compared with existing methods, our segmentation method can identify the vortex area more accurately, while the time consumption is reduced more than 50%; our classification method can reduce the time consumption by more than 90% while maintaining the same classification accuracy level.

Huapeng Wu, Jie Gui, Jun Zhang et al.

Recently, deep convolutional neural network methods have achieved an excellent performance in image superresolution (SR), but they can not be easily applied to embedded devices due to large memory cost. To solve this problem, we propose a pyramidal dense attention network (PDAN) for lightweight image super-resolution in this paper. In our method, the proposed pyramidal dense learning can gradually increase the width of the densely connected layer inside a pyramidal dense block to extract deep features efficiently. Meanwhile, the adaptive group convolution that the number of groups grows linearly with dense convolutional layers is introduced to relieve the parameter explosion. Besides, we also present a novel joint attention to capture cross-dimension interaction between the spatial dimensions and channel dimension in an efficient way for providing rich discriminative feature representations. Extensive experimental results show that our method achieves superior performance in comparison with the state-of-the-art lightweight SR methods.