公开(公告)号：US20180276120A1

公开(公告)日：2018-09-27

申请号：US15624687

申请日：2017-06-15

Applicant: Microsoft Technology Licensing, LLC

Inventor： Dimitrios VYTINIOTIS ,  Manuel Silverio da Silva COSTA ,  Kapil VASWANI ,  Matthew John PARKINSON ,  Piyus Kumar KEDIA

IPC: G06F12/02

CPC classification number: G06F12/0261 ,  G06F12/023 ,  G06F12/0253 ,  G06F12/109 ,  G06F12/145 ,  G06F2212/1016 ,  G06F2212/1044 ,  G06F2212/1052 ,  G06F2212/657

Abstract: A method of manual memory management is described. In response to detecting an access violation triggered by the use of an invalid reference to an object in a manual heap, a source of the access in a register or stack is identified. An updated reference for the object using stored mapping data is determined and used to replace the invalid reference in the source.

公开(公告)号：US20180253311A1

公开(公告)日：2018-09-06

申请号：US15615757

申请日：2017-06-06

Applicant: Microsoft Technology Licensing, LLC

Inventor： Matthew John PARKINSON ,  Manuel Silverio da Silva COSTA ,  Dimitrios VYTINIOTIS ,  Kapil VASWANI

IPC: G06F9/30 ,  G06F12/06

CPC classification number: G06F9/30123 ,  G06F9/30116

Abstract: A method of manual memory management is described which comprises enabling one or more threads to access an object created in a manual heap by storing a reference to the object in thread-local state and subsequently deleting the stored reference after accessing the object. In response to abandonment of the object, an identifier for the object and a current value of either a local counter of a thread or a global counter are stored in a delete queue and all threads are prevented from storing any further references to the object in thread-local state. Deallocation of the object only occurs when all references to the object stored in thread-local state for any threads have been deleted and a current value of the local counter for the thread or the global counter has incremented to a value that is at least a pre-defined amount more than the stored value, wherein the global counter is updated using one or more local counters.

公开(公告)号：US20240419967A1

公开(公告)日：2024-12-19

申请号：US18814438

申请日：2024-08-23

Applicant: Microsoft Technology Licensing, LLC

Inventor： Ryota TOMIOKA ,  Matthew Alastair JOHNSON ,  Daniel Stefan TARLOW ,  Samuel Alexander WEBSTER ,  Dimitrios VYTINIOTIS ,  Alexander Lloyd GAUNT ,  Maik RIECHERT

IPC: G06N3/08 ,  G06N3/04 ,  G06N3/063

Abstract: A neural network training apparatus is described which has a network of worker nodes each having a memory storing a subgraph of a neural network to be trained. The apparatus has a control node connected to the network of worker nodes. The control node is configured to send training data instances into the network to trigger parallelized message passing operations which implement a training algorithm which trains the neural network. At least some of the message passing operations asynchronously update parameters of individual subgraphs of the neural network at the individual worker nodes.

公开(公告)号：US20220222531A1

公开(公告)日：2022-07-14

申请号：US17706586

申请日：2022-03-28

Applicant: Microsoft Technology Licensing, LLC

Inventor： Ryota TOMIOKA ,  Matthew Alastair JOHNSON ,  Daniel Stefan TARLOW ,  Samuel Alexander WEBSTER ,  Dimitrios VYTINIOTIS ,  Alexander Lloyd GAUNT ,  Maik RIECHERT

IPC: G06N3/08 ,  G06N3/04 ,  G06N3/063

Abstract: A neural network training apparatus is described which has a network of worker nodes each having a memory storing a subgraph of a neural network to be trained. The apparatus has a control node connected to the network of worker nodes. The control node is configured to send training data instances into the network to trigger parallelized message passing operations which implement a training algorithm which trains the neural network. At least some of the message passing operations asynchronously update parameters of individual subgraphs of the neural network at the individual worker nodes.

公开(公告)号：US20180336458A1

公开(公告)日：2018-11-22

申请号：US15599058

申请日：2017-05-18

Applicant: Microsoft Technology Licensing, LLC

Inventor： Ryota TOMIOKA ,  Matthew Alastair JOHNSON ,  Daniel Stefan TARLOW ,  Samuel Alexander WEBSTER ,  Dimitrios VYTINIOTIS ,  Alexander Lloyd GAUNT ,  Maik RIECHERT

IPC: G06N3/08 ,  G06N3/04

CPC classification number: G06N3/063

Abstract: A neural network training apparatus is described which has a network of worker nodes each having a memory storing a subgraph of a neural network to be trained. The apparatus has a control node connected to the network of worker nodes. The control node is configured to send training data instances into the network to trigger parallelized message passing operations which implement a training algorithm which trains the neural network. At least some of the message passing operations asynchronously update parameters of individual subgraphs of the neural network at the individual worker nodes.