公开(公告)号：US20230176907A1

公开(公告)日：2023-06-08

申请号：US18074440

申请日：2022-12-02

Applicant: Apple Inc.

Inventor： Francesco ROSSI ,  Gaurav KAPOOR ,  Michael R. SIRACUSA ,  William B. MARCH

IPC: G06F9/50 ,  G06N3/063 ,  G06F8/41 ,  G06F9/48 ,  G06N3/02

CPC classification number: G06F9/50 ,  G06F9/5044 ,  G06N3/063 ,  G06F8/451 ,  G06F9/485 ,  G06F9/5038 ,  G06N3/02 ,  G06F2209/509

Abstract: The subject technology provides for dynamic task allocation for neural network models. The subject technology determines an operation performed at a node of a neural network model. The subject technology assigns an annotation to indicate whether the operation is better performed on a CPU or a GPU based at least in part on hardware capabilities of a target platform. The subject technology determines whether the neural network model includes a second layer. The subject technology, in response to determining that the neural network model includes a second layer, for each node of the second layer of the neural network model, determines a second operation performed at the node. Further the subject technology assigns a second annotation to indicate whether the second operation is better performed on the CPU or the GPU based at least in part on the hardware capabilities of the target platform.

公开(公告)号：US20250131286A1

公开(公告)日：2025-04-24

申请号：US19000562

申请日：2024-12-23

Applicant: Apple Inc.

Inventor： Gaurav KAPOOR ,  Cecile M. FORET ,  Francesco ROSSI ,  Kit-Man WAN ,  Umesh S. VAISHAMPAYAN ,  Etienne BELANGER ,  Albert ANTONY ,  Alexey MARINICHEV ,  Marco ZULIANI ,  Xiaojin SHI

IPC: G06N3/10 ,  G06F8/41 ,  G06F9/50 ,  G06N3/04 ,  G06N3/063 ,  G06N3/08

Abstract: The subject technology provides receiving a neural network (NN) model to be executed on a target platform, the NN model including multiple layers that include operations and some of the operations being executable on multiple processors of the target platform. The subject technology further sorts the operations from the multiple layers in a particular order based at least in part on grouping the operations that are executable by a particular processor of the multiple processors. The subject technology determines, based at least in part on a cost of transferring the operations between the multiple processors, an assignment of one of the multiple processors for each of the sorted operations of each of the layers in a manner that minimizes a total cost of executing the operations. Further, for each layer of the NN model, the subject technology includes an annotation to indicate the processor assigned for each of the operations.

公开(公告)号：US20200380956A1

公开(公告)日：2020-12-03

申请号：US16552309

申请日：2019-08-27

Applicant: Apple Inc.

Inventor： Francesco ROSSI ,  Sivanand ACHANTA

IPC: G10L15/16 ,  G06N3/04 ,  G10L21/14

Abstract: Systems and processes for generating speech from text are provided. An example method of generating speech from text includes, at an electronic device having at least one processor and memory, obtaining text; generating a plurality of segments of a spectrogram using a first neural network, each spectrogram segment of the plurality of spectrogram segments representing a portion of the text; generating, based on the plurality of spectrogram segments, a plurality of speech segments using a second neural network; and providing the plurality of speech segments as a speech output.

公开(公告)号：US20230177350A1

公开(公告)日：2023-06-08

申请号：US17903991

申请日：2022-09-06

Applicant: Apple Inc.

Inventor： Gaurav KAPOOR ,  Cecile M. FORET ,  Francesco ROSSI ,  Kit-Man WAN ,  Umesh S. VAISHAMPAYAN ,  Etienne BELANGER ,  Albert ANTONY ,  Alexey MARINICHEV ,  Marco ZULIANI ,  Xiaojin SHI

IPC: G06N3/10 ,  G06N3/08 ,  G06N3/04 ,  G06F8/41 ,  G06N3/063

CPC classification number: G06N3/10 ,  G06N3/08 ,  G06N3/04 ,  G06F8/443 ,  G06F8/41 ,  G06F8/4441 ,  G06N3/063 ,  G06F9/50

Abstract: The subject technology provides receiving a neural network (NN) model to be executed on a target platform, the NN model including multiple layers that include operations and some of the operations being executable on multiple processors of the target platform. The subject technology further sorts the operations from the multiple layers in a particular order based at least in part on grouping the operations that are executable by a particular processor of the multiple processors. The subject technology determines, based at least in part on a cost of transferring the operations between the multiple processors, an assignment of one of the multiple processors for each of the sorted operations of each of the layers in a manner that minimizes a total cost of executing the operations. Further, for each layer of the NN model, the subject technology includes an annotation to indicate the processor assigned for each of the operations.

公开(公告)号：US20200167193A1

公开(公告)日：2020-05-28

申请号：US16776338

申请日：2020-01-29

Applicant: Apple Inc.

Inventor： Francesco ROSSI ,  Gaurav KAPOOR ,  Michael R. SIRACUSA ,  William B. MARCH

IPC: G06F9/50 ,  G06F8/41 ,  G06N3/02 ,  G06N3/063

Abstract: The subject technology provides for dynamic task allocation for neural network models. The subject technology determines an operation performed at a node of a neural network model. The subject technology assigns an annotation to indicate whether the operation is better performed on a CPU or a GPU based at least in part on hardware capabilities of a target platform. The subject technology determines whether the neural network model includes a second layer. The subject technology, in response to determining that the neural network model includes a second layer, for each node of the second layer of the neural network model, determines a second operation performed at the node. Further the subject technology assigns a second annotation to indicate whether the second operation is better performed on the CPU or the GPU based at least in part on the hardware capabilities of the target platform.

公开(公告)号：US20210397596A1

公开(公告)日：2021-12-23

申请号：US17028920

申请日：2020-09-22

Applicant: Apple Inc.

Inventor： Albert ANTONY ,  Francesco ROSSI ,  Guillaume TARTAVEL ,  Xiaojin SHI ,  Marco ZULIANI

IPC: G06F16/22 ,  G06N3/04

Abstract: The subject technology provides a framework for evaluating activation functions of a neural network using lookup tables. In order to provide lookup table based activation functions with a desired precision within hardware constraints for the lookup tables, multiple lookup tables for each activation function can be provided. Each of the multiple lookup tables may correspond to a respective subrange of input values, within a full range of input values for the activation function.

公开(公告)号：US20200379740A1

公开(公告)日：2020-12-03

申请号：US16583191

申请日：2019-09-25

Applicant: Apple Inc.

Inventor： Timothy S. PAEK ,  Francesco ROSSI ,  Jamil DHANANI ,  Keith P. AVERY ,  Minwoo JEONG ,  Xiaojin SHI ,  Harveen KAUR ,  Brandt M. WESTING

IPC: G06F8/41 ,  G06N3/04 ,  G06N3/08 ,  G06F8/60

Abstract: The subject technology receives a neural network model in a model format, the model format including information for a set of layers of the neural network model, each layer of the set of layers including a set of respective operations. The subject technology generates neural network (NN) code from the neural network model, the NN code being in a programming language distinct from the model format, and the NN code comprising a respective memory allocation for each respective layer of the set of layers of the neural network model, where the generating comprises determining the respective memory allocation for each respective layer based at least in part on a resource constraint of a target device. The subject technology compiles the NN code into a binary format. The subject technology generates a package for deploying the compiled NN code on the target device.

公开(公告)号：US20200082274A1

公开(公告)日：2020-03-12

申请号：US16262809

申请日：2019-01-30

Applicant: Apple Inc.

Inventor： Francesco ROSSI ,  Cecile M. FORET ,  Gaurav KAPOOR ,  Kit-Man WAN ,  Umesh S. VAISHAMPAYAN ,  Etienne BELANGER ,  Albert ANTONY ,  Alexey MARINICHEV ,  Marco ZULIANI ,  Xiaojin SHI

IPC: G06N3/10 ,  G06N3/04 ,  G06N3/08

Abstract: The subject technology provides receiving a neural network (NN) model to be executed on a target platform, the NN model including multiple layers that include operations and some of the operations being executable on multiple processors of the target platform. The subject technology further sorts the operations from the multiple layers in a particular order based at least in part on grouping the operations that are executable by a particular processor of the multiple processors. The subject technology determines, based at least in part on a cost of transferring the operations between the multiple processors, an assignment of one of the multiple processors for each of the sorted operations of each of the layers in a manner that minimizes a total cost of executing the operations. Further, for each layer of the NN model, the subject technology includes an annotation to indicate the processor assigned for each of the operations.

公开(公告)号：US20200082273A1

公开(公告)日：2020-03-12

申请号：US16262807

申请日：2019-01-30

Applicant: Apple Inc.

Inventor： Francesco ROSSI ,  Cecile M. FORET ,  Gaurav KAPOOR ,  Kit-Man WAN ,  Umesh S. VAISHAMPAYAN ,  Etienne BELANGER

IPC: G06N3/10 ,  G06F9/48

Abstract: The subject technology runs a compiled neural network (NN) model on a particular processor with multiple priority queues for executing different processes, the compiled NN model being assigned to a particular priority queue, and the compiled NN model includes context switch instructions that were previously inserted into a neural network (NN) model from which the compiled NN model was compiled. The subject technology determines that a particular context switch instruction has been executed by the particular processor. The subject technology determines that a different process is waiting to be executed, the different process being assigned to a different priority queue and the different process being a higher priority process than the running compiled NN model. In response to executing the particular context switch instruction, the subject technology performs a context switch to the different process assigned to the different priority queue when the different process is waiting to be executed.