基于卷积神经网络任务分解算法的细粒度调度方法研究——以人工智能ChatGPT为例

Fine-grained scheduling method based on convolutional neural network task decomposition algorithm-Taking artificial intelligence ChatGPT as an example

在人工智能快速发展的背景下,如何有效地分解和调度复杂任务,如大规模语言模型ChatGPT,成为关键问题。研究采用了卷积神经网络设计的任务分解算法,并结合细粒度调度策略,以提升处理效率和准确性。结果显示,在教学相关大规模参数模型中,相较于其他调度方法,新方法在系统周转时间和吞吐量方面表现优异,分别降低了57.7%和21.1%,提升了252.1%和52.9%。综上可知,结合细粒度调度策略的卷积神经网络任务分解算法,能够为大规模语言模型带来更高效、更准确的处理能力。

In the context of the rapid development of artificial intelligence,how to effectively decompose and schedule complex tasks,such as the large-scale language model ChatGPT,has become a key issue.The study adopted a task decomposition algorithm designed using convolutional neural networks and combined it with fine-grained scheduling strategies to improve processing efficiency and accuracy.The results show that in large-scale parameter models related to teaching compared to other scheduling methods,the new method performs excellently in system turnover time and throughput,reducing by 57.7%and 21.1%respectively,and improving by 252.1%and 52.9%.In summary,the convolutional neural network task decomposition algorithm combined with fine-grained scheduling strategy can bring more efficient and accurate processing capabilities to large-scale language models.