Zhejiang Lab

Huawei Technologies Co., Limited

Weiqiang Jia

Hujun Bao

Hongsheng Wang

Wei Hua

Ding Li

Chunhua Tan

Liyu Wang

Wenqiang Yang

Pengli Ji

The disclosure discloses a method of neural network model computation-oriented intermediate representation and apparatus thereof. The method includes the following steps: S1, parsing an input model file so as to acquire topological structure information of a neural network; S2, constructing a logical computation graph; S21, inferring physical layout information of each operator in the logical computation graph; S22, inferring meta attributes of each operator in the logical computation graph; S23, inferring description information of input and output logical tensors of each operator in the logical computation graph; S3, constructing a physical computation graph; S31, generating a physical computation graph, etc.

Method of neural network model computation-oriented intermediate representation by constructing physical computation graph, inferring information of input and output tensor edges of each node therein, performing memory optimization on tensor edges, and optimizing physical computation graph

Disclosed is a method for distributed type training adaptation and apparatus in a deep learning framework and an AI accelerator card. The method includes the following steps: S1: the deep learning framework supports single-card configuration in a newly added AI accelerator card, and sub-steps thereof are as follows: S11: the deep learning framework supports new hardware; S12: the deep learning framework supports a device thread of the new hardware; S13: the deep learning framework supports a memory operation of the new hardware; and S14: the deep learning framework supports an operator kernel function of the new hardware; S2: the deep learning framework supports multi-card configuration in the newly added AI accelerator card; S3: the deep learning framework supports tensor segmentation and multi-card distribution; and S4: the deep learning framework supports multi-card collective communication in the newly added AI accelerator card.

Method for distributed type training adaptation and apparatus in deep learning framework and AI accelerator card

In a data prefetching method, a storage device obtains a first sequence stream length and a first access count of a target logical block after execution of a first data access request is completed. When a second data access request is received, the storage device modifies the first sequence stream length to a second sequence stream length and modifies the first access count to a second access count. The storage device further calculates a sequence degree of the target logical block based on the second sequence stream length and the second access count, and performs a data prefetching operation when the sequence degree of the target logical block exceeds a first prefetch threshold.

Data prefetching method and apparatus, and storage device

Growing community of inventors

Hangzhou, China

Weiqiang Jia