Computing device and parameter synchronization method implemented by computing device

A computing device comprising:

• a processor;
• and a memory storing a plurality of instructions, which when executed by the processor, cause the processor to:
• import a deep learning training task of a preset model into a server communicatively coupled to the computing device;
• record a preset number of iterative processes during the deep learning training;
• divide each iterative process into a plurality of phases according to time, the plurality of phases including at least an Host to Device (H2D) phase for a host to copy data to a Graphics Processing Unit (GPU) device, a Device to Host (D2H) phase for a Global Positioning System (GPS) device to copy data to the host, and a Central Processing Unit (CPU) phase for a CPU to work;
• determine whether a time ratio of the H2D phase, the D2H phase, and the CPU phase in each iterative process is greater than a preset value;
• and confirm the server to use a copy mode for performing parameter synchronization when the time ratio of the H2D, D2H, and CPU phases is determined to be greater than the preset value.