For CNNs based stereo matching methods, cost volumes play an important role in achieving good matching accuracy. In this paper, we present an end-to-end trainable convolution neural network to fully use cost volumes for stereo matching. Our network consists of three sub-modules, i.e., shared feature extraction, initial disparity estimation, and disparity refinement. These sub-modules of our network are tightly-coupled, making it compact and easy to train. Moreover, we investigate the problem of developing a robust model to perform well across multiple datasets with different characteristics. We achieve this by introducing a two-stage finetuning scheme to gently transfer the model to target datasets. Specifically, in the first stage, the model is finetuned using both a large synthetic dataset and the target datasets with a relatively large learning rate, while in the second stage the model is trained using only the target datasets with a small learning rate. The proposed method is tested on several benchmarks including the Middlebury 2014, KITTI 2015, ETH3D 2017, and SceneFlow datasets. Experimental results show that our method achieves the state-of-the-art performance on all the datasets. The proposed method also won the 1st prize on the Stereo task of Robust Vision Challenge 2018.

Foundation models are in the process of becoming the dominant deep learning technology. Pretraining a foundation model is always time-consuming due to the large scale of both the model parameter and training dataset. Besides being computing-intensive, the pretraining process is extremely memory- and communication-intensive. These challenges make it necessary to apply 3D parallelism, which integrates data parallelism, pipeline model parallelism, and tensor model parallelism, to achieve high training efficiency. However, current 3D parallelism frameworks still encounter two issues: i) they are not transparent to model developers, requiring manual model modification to parallelize training, and ii) their utilization of computation resources, GPU memory, and network bandwidth is insufficient. We propose Merak , an automated 3D parallelism deep learning training framework with high resource utilization. Merak automatically deploys 3D parallelism with an automatic model partitioner, which includes a graph-sharding algorithm and proxy node-based model graph. Merak also offers a non-intrusive API to scale out foundation model training with minimal code modification. In addition, we design a high-performance 3D parallel runtime engine that employs several techniques to exploit available training resources, including a shifted critical path pipeline schedule that increases computation utilization, stage-aware recomputation that makes use of idle worker memory, and sub-pipelined tensor model parallelism that overlaps communication and computation. Experiments on 64 GPUs demonstrate Merak's capability to speed up training performance over state-of-the-art 3D parallelism frameworks of models with 1.5, 2.5, 8.3, and 20 billion parameters by up to 1.42, 1.39, 1.43, and 1.61×, respectively.

Due to the capacity of storing massive data and providing huge computing resources, cloud computing has been a desirable platform for doing machine learning. However, the issue of data privacy is far from being well solved and thus has been a general concern in the cloud-aided machine learning. In this work, we investigate the study of how to efficiently do decision tree training and evaluation in the cloud and meanwhile achieve privacy preservation. Unlike existing cloud server-assisted model training approaches, in our proposed solution, the whole training process is mostly done by the cloud service provider who owns the machine learning model. Since the cloud cannot directly divide the encrypted dataset according to the best attributes selected, we propose a new method for decision tree training without dataset splitting. Precisely, we design three methods for decision tree training with the different tradeoff between privacy and efficiency. In all of these methods, the outsourced data are not revealed to the cloud service provider. We also propose a privacy-preserving decision tree evaluation scheme where the cloud service provider learns nothing about the user’s input and the classification result while the trained model is kept secret to the user who could only learn the classification result. Compared with previous decision tree evaluation work, our scheme achieves desirable privacy preservation against both the user and the cloud service provider, and also minimizes the user’s computation and communication costs. Moreover, besides protecting the data confidentiality, our proposed scheme also supports off-line users and thus has good scalability. The real-world dataset-based experimental results demonstrate that our system is of desirable utility and efficiency.

Yifu Gao, Linbo Qiao, Zhigang Kan, Zhihua Wen, Yongquan He, Dongsheng Li. Findings of the Association for Computational Linguistics ACL 2024. 2024.

Deep learning (DL) has gained great success in recent years, leading to state-of-the-art performance in research community and industrial fields like computer vision and natural language processing. One of the reasons for this success is the huge amount parameters adopted in DL models. However, it is impractical to train a moderately large model with a large number of parameters on a typical single device. Thus, It is necessary to train DL models in clusters with distributed training algorithms. However, traditional distributed training algorithms are usually sub-optimal and highly customized, which owns the drawbacks to train large-scale DL models in varying computing clusters. To handle the above problem, researchers propose auto-parallelism, which is promising to train large-scale DL models efficiently and practically in various computing clusters. In this survey, we perform a broad and thorough investigation on challenges, basis, and strategy searching methods of auto-parallelism in DL training. First, we abstract basic parallelism schemes with their communication cost and memory consumption in DL training. Further, we analyze and compare a series of current auto-parallelism works and investigate strategies and searching methods which are commonly used in practice. At last, we discuss several trends in auto-parallelism which are promising in further research.