Beijing Baidu Netcom Science and Technology Co., Ltd.

Kunlunxin Technology (beijing) Company Limited

Baidu USA LLC

Baidu.com Times Technology (beijing) Co. Ltd.

Vmware, Inc.

Kunlunxin Technology (beiiing) Company Limited

Jian Ouyang

Wei Qi

Yueqiang Cheng

Yong Liu

Tao Wei

Yong Wang

Jing Wang

Hefei Zhu

Zhibiao Zhao

Zihao Liang

Peng Wu

Xiaozhang Gong

Qingshu Chen

Ningyi Xu

Huimin Li

Yichen Tu

Ni Zhou

Liang Gao

Zhao Zhang

Lin Liu

Yupeng Li

Kecheng Lu

Xueliang Du

Yingnan Xu

Canghai Gu

Sijie Yang

James Kiryakoza

Kang An

Jun Huan

Jiaxin Shi

Yingzhen Yang

Baoxin Zhao

Sizhong Li

Chongqin Wang

Chongquin Wang

A method for processing an image, an electronic device, and a storage medium are provided. The method includes: acquiring an input image, the input image including a plurality of channels, and each channel of the plurality of channels including a plurality of pixel points; performing pixel extractions in parallel using a plurality of dedicated processing units on each channel of the input image, to obtain pixel point data for each corresponding pixel point of each channel; and splicing the pixel point data for each corresponding pixel point of each channel to obtain an output image.

Method for processing image, electronic device and storage medium

The present disclosure provides a method of processing an image, an electronic device, and a storage medium, which may be used in a field of artificial intelligence, especially in a field of image processing, etc. The method includes: acquiring an input image containing a plurality of rows of pixels; performing, by using a plurality of dedicated processing units, a pixel extraction in parallel on each row of pixels of the plurality of rows of pixels in the input image, so as to obtain row data for each row of pixels; and stitching the row data for each row of pixels, so as to obtain an output image.

Method of processing image, electronic device, storage medium, and program product

Embodiments of the present disclosure propose a deep learning processing apparatus and method, device and storage medium, relating to the field of artificial intelligence. A deep learning processing apparatus includes: at least one matrix multiply-add module, configured to perform a matrix multiply-add operation of a convolution kernel parameter value matrix of a convolutional layer in a convolutional neural network and a first error gradient value matrix to obtain a plurality of intermediate matrices; a storage apparatus, configured to store the plurality of intermediate matrices without reshaping elements in the plurality of intermediate matrices; and a plurality of matrix accumulation modules, configured to read the plurality of intermediate matrices from the storage apparatus and perform a matrix accumulation operation based on the plurality of intermediate matrices according to a convolution scheme of the convolutional layer in parallel, to obtain a second error gradient value matrix for the convolutional layer.

Deep learning processing apparatus and method, device and storage medium

Presented herein are embodiments of an improved asymmetric quantization, which may generally be referred to as improved asymmetric quantization (IAQ) embodiments. IAQ embodiments combine the benefits of conventional asymmetric quantization and symmetric quantization but also provide additional computation efficiencies. Embodiments of IAQ adopt an asymmetric range of the weights of a neural network layer, so they circumvent the limitation of symmetric range of symmetric quantization. Moreover, the inference process of a neural network quantized by an IAQ embodiment is much faster than that of the neural network quantized by conventional asymmetric quantization by quantizing an offset value of each layer.

Asymmetric quantization for compression and for acceleration of inference for neural networks

According to one embodiment, a DP accelerator includes one or more execution units (EUs) configured to perform data processing operations in response to an instruction received from a host system coupled over a bus. The DP accelerator includes a time unit (TU) coupled to the security unit to provide timestamp services. The DP accelerator includes a security unit (SU) configured to establish and maintain a secure channel with the host system to exchange commands and data associated with the data processing operations, where the security unit includes a secure storage area to store a private root key associated with the DP accelerator, where the private root key is utilized for authentication. The SU includes a random number generator to generate a random number, and a cryptographic engine to perform cryptographic operations on data exchanged with the host system over the bus using a session key derived based on the random number.

Data processing accelerator having a security unit to provide root trust services

Example embodiments of the present application provide an instruction executing method and apparatus, an electronic device, and a computer-readable storage medium that may be applied in the field of artificial intelligence. The instruction executing method may include: executing an instruction sequence that includes memory instructions and non-memory instructions, the instructions in the sequence executed starting to be executed in order; determining that execution of a first memory instruction needs to be completed before a second memory instruction starts to be executed, the second memory instruction being a next memory instruction following the first memory instruction in the instruction sequence; and executing non-memory instructions between the first memory instruction and the second memory instruction without executing the second memory instruction, during a cycle of executing the first memory instruction.

Instruction executing method and apparatus, electronic device, and computer-readable storage medium

According to one embodiment, a system performs a secure boot using a security module such as a trusted platform module (TPM) of a host system. The system establishes a trusted execution environment (TEE) associated with one or more processors of the host system. The system launches a memory manager within the TEE, where the memory manager is configured to manage memory resources of a data processing (DP) accelerator coupled to the host system over a bus, including maintaining memory usage information of global memory of the DP accelerator. In response to a request received from an application running within the TEE for accessing a memory location of the DP accelerator, the system allows or denies the request based on the memory usage information.

Method and system for managing memory of data processing accelerators

Embodiments of the present disclosure provide a data accessing method, a device and a storage medium. The method includes: obtaining a first accessing request and a second accessing request for a storage device; loading first data associated with the first accessing request from a source device to a pre-allocated buffer area with a size same as a size of a single physical storage block of the storage device; determining a first part of the second data when the first size of second data associated with the second accessing request is greater than or equal to the second size of an available space of the buffer area, a size of the first part being the same as the second size; and providing the first data and the first part to a target device associated with the first accessing request and the second accessing request.

Data accessing method, device, and storage medium

A data processing system includes a central processing unit (CPU) and accelerator cards coupled to the CPU over a bus, each of the accelerator cards having a plurality of data processing (DP) accelerators to receive DP tasks from the CPU and to perform the received DP tasks. At least two of the accelerator cards are coupled to each other via an inter-card connection, and at least two of the DP accelerators are coupled to each other via an inter-chip connection. Each of the inter-card connection and the inter-chip connection is capable of being dynamically activated or deactivated, such that in response to a request received from the CPU, any one of the accelerator cards or any one of the DP accelerators within any one of the accelerator cards can be enabled or disabled to process any one of the DP tasks received from the CPU.

Distributed AI training topology based on flexible cable connection

According to one embodiment, a system receives, at a host channel manager (HCM) of a host system, a request from an application to establish a secure channel with a data processing (DP) accelerator, where the DP accelerator is coupled to the host system over a bus. In response to the request, the system generates a first session key for the secure channel based on a first private key of a first key pair associated with the HCM and a second public key of a second key pair associated with the DP accelerator. In response to a first data associated with the application to be sent to the DP accelerator, the system encrypts the first data using the first session key. The system then transmits the encrypted first data to the DP accelerator via the secure channel over the bus.

Method for establishing a secure information exchange channel between a host system and a data processing accelerator

According to one embodiment, a data processing system performs a secure boot using a security module (e.g., a trusted platform module (TPM)) of a host system. The system verifies that an operating system (OS) and one or more drivers including an accelerator driver associated with a data processing (DP) accelerator is provided by a trusted source. The system launches the accelerator driver within the OS. The system generates a trusted execution environment (TEE) associated with one or more processors of the host system. The system launches an application and a runtime library within the TEE, where the application communicates with the DP accelerator via the runtime library and the accelerator driver.

Method and system for protecting data processed by data processing accelerators

According to various embodiments, methods and systems are provided to accelerate artificial intelligence (AI) model training with advanced interconnect communication technologies and systematic zero-value compression over a distributed training system. According to an exemplary method, during each iteration of a Scatter-Reduce process performed on a cluster of processors arranged in a logical ring to train a neural network model, a processor receives a compressed data block from a prior processor in the logical ring, performs an operation on the received compressed data block and a compressed data block generated on the processor to obtain a calculated data block, and sends the calculated data block to a following processor in the logical ring. A compressed data block calculated from corresponding data blocks from the processors can be identified on each processor and distributed to each other processor and decompressed therein for use in the AI model training.

Accelerating AI training by an all-reduce process with compression over a distributed system

An automatic processing system, a system on chip and a method for monitoring a processing module are described herein. The automatic driving processing system comprises: an automatic driving processing module, configured for receiving an input data stream and processing the input data stream based on a deep learning model so as to generate a processing result; a fault detection module, configured for generating a control signal and a fault detection stimulating data stream, and receiving the processing result from the automatic driving processing module; and a multi-way selection module, configured for receiving an automatic driving data stream as well as the control signal and the fault detection stimulating data stream, and selectively outputting the automatic driving data stream or the fault detection stimulating data stream to the automatic driving processing module based on the control signal, as an input data stream.

Automatic driving processing system, system on chip and method for monitoring processing module

Embodiments of the present disclosure provide a method and apparatus for extracting image data in parallel from multiple convolution windows, a device, and a computer-readable storage medium. The method includes: dividing an image into multiple groups of convolution windows, where the multiple groups of convolution windows include a first group of convolution windows and a second group of convolution windows, and each group of convolution windows include multiple convolution windows. The method further includes extracting image data in parallel from multiple convolution windows in the first group of convolution windows by using multiple data processing units, and extracting, after the extraction of image data from the first group of convolution windows is completed, image data from multiple convolution windows in the second group of convolution windows in parallel by using the multiple data processing units.

Method and apparatus for extracting image data in parallel from multiple convolution windows, device, and computer-readable storage medium

According to one embodiment, a system receives, at a host system a public attestation key (PK_ATT) or a signed PK_ATT from a data processing (DP) accelerator over a bus. The system verifies the PK_ATT using a public root key (PK_RK) associated with the DP accelerator. In response to successfully verifying the PK_ATT, the system transmits a kernel identifier (ID) to the DP accelerator to request attesting a kernel object stored in the DP accelerator. In response to the system receives a kernel digest or a signed kernel digest corresponding to the kernel object from the DP accelerator, verifying the kernel digest using the PK_ATT. The system sends the verification results to the DP accelerator for the DP accelerator to access the kernel object based on the verification results.

Growing community of inventors

Beijing, China

Jian Ouyang