Back to results
US 11,455,232 B2Granted
Debug operations on artificial intelligence operations
AssigneeMicron Technology, Inc.
InventorAlberto Troia
Life of this patent
Estimating the term…
Abstract
The present disclosure includes apparatuses and methods related to performing a debug operation on an artificial intelligence operation. An example apparatus can include a number of memory arrays and a controller, wherein the controller is configured to perform an artificial intelligence (AI) operation on data stored in the number of memory arrays and perform a debug operation on the AI operation.
Claims22
An apparatus, comprising:
- a number of memory arrays that are a number of layers of a neural network;
- and
- a controller, wherein the controller is configured to cause the apparatus to:
- perform an artificial intelligence (AI) operation on data stored in the number of memory arrays;
- perform a debug operation on the AI operation to identify errors of the AI operation;
- stop the AI operation in response to performing the debug operation;
- store data including the identified errors of the AI operation in a temporary memory block of the number of memory arrays in response to a particular bit of a first register being programmed to a first state;
- define, via a second register, a layer of the number of layers of the neural network where errors consistently occur at least partially based on the identified errors;
- change the data stored in the temporary memory block to correct the identified errors;
- and
- continue the AI operation using the changed data.
8 more elements
An apparatus, comprising:
- a number of memory arrays that are a number of layers of a neural network;
- and
- a controller, wherein the controller is configured to:
- receive, from a host, a debug command to perform a debug operation on an artificial intelligence (AI) operation performed on the apparatus;
- execute the debug command by causing performance of the debug operation on the AI operation to identify errors of the AI operation;
- stop the AI operation in response to executing the debug command;
- store data including the identified errors of the AI operation in a temporary memory block of the number of memory arrays in response to a particular bit of a first register being programmed to a first state;
- define, via a second register, a layer of the number of layers of the neural network where errors consistently occur at least partially based on the identified errors;
- change the data stored in the temporary memory block to correct the identified errors;
- and
- continue the AI operation using the changed data.
8 more elements
An apparatus, comprising:
- a number of memory arrays that are a number of layers of a neural network;
- and
- a controller, wherein the controller is configured to cause the apparatus to:
- enable performance of a debug operation on an artificial intelligence (AI) operation to identify errors of the AI operation by writing an indicator in a first register to a particular state in response to receiving a command from a host;
- stop the AI operation in response to the performance of the debug operation;
- store data including the identified errors of the AI operation in a temporary memory block of the number of memory arrays in response to writing an indicator in a second register to a different particular state;
- define, via a third register, a layer of the number of layers of the neural network where errors consistently occur at least partially based on the identified errors;
- change the data stored in the temporary memory block to correct the identified errors;
- and
- continue the AI operation using the changed data.
7 more elements
A method, comprising:
- performing an artificial intelligence (AI) operation on data stored in a number of memory arrays that are a number of layers of a neural network;
- performing a debug operation on the AI operation to identify errors of the AI operation;
- stopping the AI operation in response to performing the debug operation;
- storing data including the identified errors of the AI operation in temporary memory blocks in response to a particular bit of a first register being programmed to a first state;
- defining, via a second register, a layer of the number of layers of the neural network where errors consistently occur at least partially based on the identified errors;
- changing the data stored in the temporary memory blocks to correct the identified errors;
- and
- continuing the AI operation using the changed data.
5 more elements
Dependencies are inferred from the claim text.