Job packing is an effective technique to harvest the idle resources allocated to the deep learning (DL) training jobs but not fully utilized, especially when clusters may experience low utilization, and users may overestimate their resource needs. However, existing job packing techniques tend to be conservative due to the mismatch in scope and granularity between job packing and cluster scheduling. In particular, tapping the potential of job packing in the training cluster requires a local and fine-grained coordination mechanism. To this end, we propose a novel job-packing middleware named Gimbal , which operates between the cluster scheduler and the hardware resources. As middleware, Gimbal must not only facilitate coordination among the packed jobs but also support various scheduling objectives of different schedulers. Gimbal achieves dual functionality by introducing a set of worker calibration primitives designed to calibrate workers’ execution status in a fine-grained manner. The primitives obscure the complexity of the underlying job and resource management mechanisms, thus offering the generality and extensibility for crafting coordination policies tailored to various scheduling objectives. We implement Gimbal on a real-world GPU cluster and evaluate it with a set of representative DL training jobs. The results show that Gimbal improves different scheduling objectives up to 1.32× compared with the state-of-the-art job packing techniques.

Abstract Camouflaged object detection (COD) aims to segment objects embedded in the environment from the background. Most existing methods are easily affected by background interference in cluttered environments and cannot accurately locate camouflage areas, resulting in over‐segmentation or incomplete segmentation structures. To effectively improve the performance of COD, we propose a spatial perceptual activation network (SPANet). SPANet extracts the spatial positional relationship between each object in the scene by activating spatial perception and uses it as global information to guide segmentation. It mainly consists of three modules: perceptual activation module (PAM), feature inference module (FIM), and interaction recovery module (IRM). Specifically, the authors design a PAM to model the positional relationship between the camouflaged object and the surrounding environment to obtain semantic correlation information. Then, a FIM that can effectively combine correlation information to suppress background interference and re‐encode to generate multi‐scale features is proposed. In addition, to further fuse multi‐scale features, an IRM to mine the complementary information and differences between features at different scales is designed. Extensive experimental results on four widely used benchmark datasets (i.e. CAMO, CHAMELEON, COD10K, and NC4K) show that the authors’ method outperforms 13 state‐of‐the‐art methods.

In large-scale distributed linear transform problems, coded computation plays an important role to reduce the delay caused by slow machines. However, existing coded schemes could end up destroying the significant sparsity that exists in large-scale machine learning problems, and in turn increase the computational delay. In this paper, we propose a coded computation strategy, referred to as diagonal code, that achieves the optimum recovery threshold and the optimum computation load. Furthermore, by leveraging the ideas from random proposal graph theory, we design a random code that achieves a constant computation load, which significantly outperforms the existing best known result. We apply our schemes to the distributed gradient descent problem and demonstrate the advantage of the approach over current fastest coded schemes.