XPipe: Efficient Pipeline Model Parallelism for Multi-GPU DNN Training
This addresses the bottleneck of GPU underutilization in distributed deep learning training, offering a practical improvement for researchers and engineers working with large models.
The paper tackles the problem of inefficient multi-GPU DNN training by proposing XPipe, an asynchronous pipeline model parallelism approach that improves GPU utilization and throughput while maintaining model accuracy comparable to synchronous methods, achieving higher throughput than state-of-the-art approaches.
We propose XPipe, an efficient asynchronous pipeline model parallelism approach for multi-GPU DNN training. XPipe is designed to use multiple GPUs to concurrently and continuously train different parts of a DNN model. To improve GPU utilization and achieve high throughput, it splits a mini-batch into a set of micro-batches. It allows the overlapping of the pipelines of multiple micro-batches, including those belonging to different mini-batches. Most importantly, the novel weight prediction strategy adopted by XPipe enables it to effectively address the weight inconsistency and staleness issues incurred by the asynchronous pipeline parallelism. As a result, XPipe incorporates the advantages of both synchronous and asynchronous pipeline model parallelism approaches. Concretely, it can achieve very comparable (even slightly better) model accuracy as its synchronous counterpart while obtaining higher throughput than it. Experimental results show that XPipe outperforms other state-of-the-art synchronous and asynchronous model parallelism approaches.