Qualcomm Incorporated

Lakshmi Narayana Macha

Prajakt Kulkarni

Haoping Xu

Zeeshan Asad Sardar

Andy I-Fong Yu

Honghao Liu

Andrei Tudorancea

Various embodiments include methods and devices for compression and decompression of weight data sets. Some embodiments may include compressing weight data by receiving a weight data set of binary numbers representing weight values, generating a frame payload including a compressed first frame of a first subset of the weight values in the weight data set, and generating a block of compressed weight data having the frame payload. Some embodiments may include decompressing weight data by retrieving a block of compressed weight data, in which the block of compressed weight data includes a frame header associated with a frame payload, in which the frame header includes a normalization factor indicator, and in which the frame payload includes compressed weight values, and generating a first decompressed frame comprising decompressed weight values of the compressed weight values of the frame payload.

Compression technique for deep neural network weights

Certain aspects of the present disclosure relate to techniques for processing (e.g., such as tasks for wireless communications and/or multimedia). According to certain aspects, a method for low power low latency data exchange generally includes providing timing information from a first processing entity to a second processing entity indicating when the second processing entity is to next process a message exchange between the first and second entities using a shared memory space, taking action to exit a low power state, based on the timing information, to access the shared memory space to process the message exchange, and entering the low power state after processing the message exchange.

Low power low latency protocol for data exchange

Growing community of inventors

San Diego, CA, United States of America

Lakshmi Narayana Macha