Advanced New Technologies Co., Ltd.

Alibaba Group Holding Limited

Advantageous New Technologies Co., Ltd.

Jianguo Xu

Guozhen Pan

Yongchao Liu

Qiyin Huang

Haitao Zhang

Yichen Tu

Ni Zhou

Guanyin Zhu

Sizhong Li

Ying Yan

Changzheng Wei

Huabing Du

Lin Wang

Boran Zhao

Xuyang Song

Implementations of the present specification provide a model-based prediction method and apparatus. The method includes: a model running environment receives an input tensor of a machine learning model; the model running environment sends a table query request to an embedding running environment, the table query request including the input tensor, to request low-dimensional conversion of the input tensor; the model running environment receives a table query result returned by the embedding running environment, the table query result being obtained by the embedding running environment by performing embedding query and processing based on the input tensor; and the model running environment inputs the table query result into the machine learning model, and runs the machine learning model to complete model-based prediction.

Separate deployment of machine learning model and associated embedding

Embodiments of the present specification provide chips and chip-based data processing methods. In an embodiment, a method comprises: obtaining data associated with one or more neural networks transmitted from a server; for each layer of a neural network of the one or more neural networks, configuring, based on the data, a plurality of operator units based on a type of computation each operator unit performs; and invoking the plurality of operator units to perform computations, based on neurons of a layer of the neural network immediately above, of the data for each neuron to produce a value of the neuron.

Chip and chip-based data processing method

An FPGA hardware device obtains encrypted data of each participant of a secure computing system, where the FPGA hardware device stores at least one first key, where the at least one first key is at least one first key of all participants in the secure computing system or at least one first key of a predetermined number of trusted managers in the secure computing system, where the FPGA hardware device includes an FPGA chip. The FPGA hardware device decrypts the encrypted data of each participant by using a working key of each participant, to obtain plaintext data of each participant, where the working key of each participant is obtained based on a corresponding first key of the at least one first key. The FPGA hardware device performs computing based on the plaintext data of each participant to obtain a computing result. The FPGA hardware device outputs the computing result.

FPGA hardware-based secure computing method and apparatus

Disclosed herein are methods, systems, and apparatus, including computer programs encoded on computer storage media, for configuring a field programmable gate array (FPGA) based trusted execution environment (TEE) for use in a blockchain network. One of the methods includes storing a device identifier (ID), a first random number, and a first encryption key in a field programmable gate array (FPGA) device; sending an encrypted bitstream to the FPGA device, wherein the encrypted bitstream can be decrypted by the first key into a decrypted bitstream comprising a second random number; receiving an encrypted message from the FPGA device; decrypting the encrypted message from the FPGA device using a third key to produce a decrypted message; in response to decrypting the encrypted message: determining a third random number in the decrypted message; encrypting keys using the third random number; and sending the keys to the FPGA device.

Field-programmable gate array based trusted execution environment for use in a blockchain network

A computer-implemented method includes obtaining a trained convolutional neural network comprising one or more convolutional layers, each of the one or more convolutional layers comprising a plurality of filters with known filter parameters; pre-computing a reusable factor for each of the one or more convolutional layers based on the known filter parameters of the trained convolutional neural network; receiving input data to the trained convolutional neural network; computing an output of the each of the one or more convolutional layers using a Winograd convolutional operator based on the pre-computed reusable factor and the input data; and determining output data of the trained convolutional network based on the output of the each of the one or more convolutional layers.

Fast computation of a convolutional neural network

Growing community of inventors

Hangzhou, China

Jianguo Xu