公开(公告)号：US11662986B1

公开(公告)日：2023-05-30

申请号：US16825282

申请日：2020-03-20

Applicant: Meta Platforms, Inc.

Inventor： Garret Ray Catron ,  Jordan Samuel Fix ,  Bertrand Allen Maher ,  Nicholas Gibson ,  Nadathur Rajagopalan Satish ,  Roman Dzhabarov ,  Hector Yuen

IPC: G06F8/41 ,  G06T1/20

CPC classification number: G06F8/41 ,  G06T1/20

Abstract: A computer program compiled for a machine learning accelerator hardware and associated with a default input data size is received. An execution of an operation of the computer program is initiated. It is identified that a data size of an input data of the operation is smaller than the default input data size. The smaller data size of the input data of the operation rather than the default input data size is caused to be transferred to the machine learning accelerator hardware for the input data of the operation.

公开(公告)号：US20240054384A1

公开(公告)日：2024-02-15

申请号：US16945679

申请日：2020-07-31

Applicant: Meta Platforms, Inc.

Inventor： Garret Ray Catron ,  Man Wang ,  Nadathur Rajagopalan Satish ,  Michael Anderson ,  Ying Zhang ,  Bertrand Allen Maher

IPC: G06N20/00 ,  G06F9/52 ,  G06F9/50

CPC classification number: G06N20/00 ,  G06F9/52 ,  G06F9/5061

Abstract: A machine learning model network is analyzed to identify types of operations and dependencies associated with different portions of the machine learning model network, including by classifying at least a portion of the types of operations as being memory bandwidth intensive or compute intensive. The machine learning model network is partitioned across a plurality of different machine learning accelerator hardware units based at least in part on the analysis. Parallelization and pipelining of an execution of the machine learning model network is allowed based on the partitioning.