Malleable fabric attached virtual artificial intelligence (ai) training appliances

Abstract

An apparatus for training artificial intelligence (AI) models is presented. In embodiments, the apparatus may include an input interface to receive in real time model training data from one or more sources to train one or more artificial neural networks (ANNs) associated with the one or more sources, each of the one or more sources associated with at least one of the ANNs; a load distributor coupled to the input interface to distribute in real time the model training data for the one or more ANNs to one or more AI appliances; and a resource manager coupled to the load distributor to dynamically assign one or more computing resources on ones of the AI appliances to each of the ANNs in view of amounts of the training data received in real time from the one or more sources for their associated ANNs.

1 - 20 . (canceled) 
     
     
         21 . One or more of a volatile memory, a non-volatile memory, or a media disc comprising instructions that, when executed, cause at least one processor to at least:
 obtain data utilized by an artificial intelligence (AI) model, the data to be executed by the AI model within an amount of time based on a service level agreement;   assign a first number of nodes to execute the model based on the data;   detect a change in model data traffic; and   add, in response to an increase in the model data traffic, a second number of nodes assigned to execute the model.   
     
     
         22 . The one or more of the volatile memory, non-volatile memory, or media disc of  claim 21 , wherein a node from the first or second number of nodes includes one or more processors and an amount of memory. 
     
     
         23 . The one or more of the volatile memory, non-volatile memory, or media disc of  claim 21 , wherein the first number or the second number is two or more. 
     
     
         24 . The one or more of the volatile memory, non-volatile memory, or media disc of  claim 23 , wherein:
 a first assigned node is located in a first zone; and   a second assigned node is located in a second zone.   
     
     
         25 . The one or more of the volatile memory, non-volatile memory, or media disc of  claim 24 , wherein the first zone and the second zone are in a same region. 
     
     
         26 . The one or more of the volatile memory, non-volatile memory, or media disc of  claim 21 , wherein to execute the model, the at least one processor is to train the AI model in a distributed architecture. 
     
     
         27 . The one or more of the volatile memory, non-volatile memory, or media disc of  claim 21 , wherein the instructions, when executed, cause the at least one processor to remove, in response to a decrease in the model data traffic, a third number of nodes assigned to execute the model. 
     
     
         28 . The one or more of the volatile memory, non-volatile memory, or media disc of  claim 21 , wherein the instructions, when executed, cause the at least one processor to obtain the data in real time. 
     
     
         29 . The one or more of the volatile memory, non-volatile memory, or media disc of  claim 21 , wherein the data is model training data. 
     
     
         30 . The one or more of the volatile memory, non-volatile memory, or media disc of  claim 21 , wherein the instructions, when executed, cause the at least one processor to:
 predict a pattern of future model data traffic; and   detect the change in the model data traffic based on the pattern.   
     
     
         31 . A method comprising:
 obtaining data utilized by an artificial intelligence (AI) model, the data to be executed by the AI model within an amount of time based on a service level agreement;   assigning a first number of nodes to execute the model based on the data;   detecting a change in model data traffic; and   adding, in response to an increase in the model data traffic, a second number of nodes assigned to execute the model.   
     
     
         32 . The method of  claim 31 , wherein a node from the first or second number of nodes includes one or more processors and an amount of memory. 
     
     
         33 . The method of  claim 31 , wherein the first number or the second number is two or more. 
     
     
         34 . The method of  claim 33 , wherein:
 a first assigned node is located in a first zone; and   a second assigned node is located in a second zone.   
     
     
         35 . The method of  claim 34 , wherein the first zone and the second zone are in a same region. 
     
     
         36 . An apparatus comprising:
 accumulator circuitry to obtain data utilized by an artificial intelligence (AI) model, the data to be executed by the AI model within an amount of time based on a service level agreement;   distributor circuitry to assign a first number of nodes to execute the model based on the data;   traffic predictor circuitry to detect a change in model data traffic; and   the distributor circuitry to add, in response to an increase in the model data traffic, a second number of nodes assigned to execute the model.   
     
     
         37 . The apparatus of  claim 36 , wherein a node from the first or second number of nodes includes one or more processors and an amount of memory. 
     
     
         38 . The apparatus of  claim 36 , wherein the first number or the second number is two or more. 
     
     
         39 . The apparatus of  claim 38 , wherein:
 a first assigned node is located in a first zone; and   a second assigned node is located in a second zone.   
     
     
         40 . The apparatus of  claim 39 , wherein the first zone and the second zone are in a same region. 
     
     
         41 . An apparatus comprising:
 means for receiving to receive data utilized by an artificial intelligence (AI) model, the data to be executed by the AI model within an amount of time based on a service level agreement;   means for distributing to assign a first number of nodes to execute the model based on the data;   means for predicting to detect a change in model data traffic; and   the means for distributing to add, in response to an increase in the model data traffic, a second number of nodes assigned to execute the model.   
     
     
         42 . The apparatus of  claim 41 , wherein a node from the first or second number of nodes includes one or more processors and an amount of memory. 
     
     
         43 . The apparatus of  claim 41 , wherein the first number or the second number is two or more. 
     
     
         44 . The apparatus of  claim 43 , wherein:
 a first assigned node is located in a first zone; and   a second assigned node is located in a second zone.   
     
     
         45 . The apparatus of  claim 44 , wherein the first zone and the second zone are in a same region.