Chia-Lin Yang

Zhi-Lin Ke, Hsiang-Yun Cheng, Chia-Lin Yang

Machine learning algorithms, such as Support Vector Machine (SVM) and Deep Neural Network (DNN), have gained a lot of interests recently. When training a machine learning algorithm, randomly shuffle all the training data can improve the testing accuracy and boost the convergence rate. Nevertheless, realizing training data random shuffling in a real system is not a straightforward process due to the slow random accesses in hard disk drive (HDD). To avoid frequent random disk access, the effect of random shuffling is often limited in existing approaches. With the emerging non-volatile memory-based storage device, such as Intel Optane SSD, which provides fast random accesses, we propose a lightweight implementation of random shuffling (LIRS) to randomly shuffle the indexes of the entire training dataset, and the selected training instances are directly accessed from the storage and packed into batches. Experimental results show that LIRS can reduce the total training time of SVM and DNN by 49.9% and 43.5% on average, and improve the final testing accuracy on DNN by 1.01%.

Chi-Chung Chen, Chia-Lin Yang, Hsiang-Yun Cheng

The training process of Deep Neural Network (DNN) is compute-intensive, often taking days to weeks to train a DNN model. Therefore, parallel execution of DNN training on GPUs is a widely adopted approach to speed up the process nowadays. Due to the implementation simplicity, data parallelism is currently the most commonly used parallelization method. Nonetheless, data parallelism suffers from excessive inter-GPU communication overhead due to frequent weight synchronization among GPUs. Another approach is pipelined model parallelism, which partitions a DNN model among GPUs, and processes multiple mini-batches concurrently. This approach can significantly reduce inter-GPU communication cost compared to data parallelism. However, pipelined model parallelism faces the weight staleness issue; that is, gradients are computed with stale weights, leading to training instability and accuracy loss. In this paper, we present a pipelined model parallel execution method that enables high GPU utilization while maintaining robust training accuracy via a novel weight prediction technique, SpecTrain. Experimental results show that our proposal achieves up to 8.91x speedup compared to data parallelism on a 4-GPU platform while maintaining comparable model accuracy.