MODEL AUTOMATIC COMPRESSION METHOD AND DEVICE FOR DEEP-LEARNING MODEL SERVING OPTIMIZATION, AND METHOD FOR PROVIDING CLOUD INFERENCE SERVICE USING SAME

摘要

The present invention relates to a model automatic compression method and device for deep-learning model serving optimization, and a method for providing a cloud inference service using same, wherein the device includes the steps of: receiving a deep-learning algorithm for constructing deep-learning models; dividing the deep-learning algorithm into a plurality of operation steps; determining at least one branch point present between the plurality of operation steps during a training process performed according to the deep-learning algorithm; generating at least one intermediate deep-learning model that branches off from the direction of progress of the training process on the basis of the at least one branch point and progresses to the final operation step of the deep-learning algorithm; and completing the deep-learning models and the at least one intermediate deep-learning model upon finishing the training process.