Dynamic allocation and re-allocation of learning model computing resources
This disclosure describes techniques for improving allocation of computing resources to computation of machine learning tasks, including on massive computing systems hosting machine learning models. A method includes a computing system, based on a computational metric trend and/or a predicted computational metric of a past task model, allocating a computing resource for computing of a machine learning task by a current task model prior to runtime of the current task model; computing the machine learning task by executing a copy of the current task model; quantifying a computational metric of the copy of the current task model; determining a computational metric trend based on the computational metric; deriving a predicted computational metric of the copy of the current task model based on the computational metric; and, based on the computational metric trend, changing allocation of a computing resource for computing of the machine learning task by the current task model.
1. A system comprising:
one or more processors; and
one or more non-transitory computer-readable media storing computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to:
compute a machine learning task by executing an additional copy of a current task model, wherein parameters of the additional copy of the current task model are learned through training of the additional copy of the current task model;
quantify a computational metric of the copy of the current task model during runtime of the additional copy of the current task model; and
change allocation of a computing resource for computing of the machine learning task by the current task model during runtime of the additional copy of the current task model.
2. The system of claim 1 , wherein the computational metric comprises a sparsity measure.
3. The system of claim 1 , wherein the computational metric comprises a learning slope.
4. The system of claim 1 , wherein the instructions further cause the one or more processors to determine a computational metric trend based on the computational metric.
5. The system of claim 1 , wherein the instructions further cause the one or more processors to derive a predicted computational metric of the additional copy of the current task model based on the computational metric.
6. The system of claim 5 , wherein deriving a predicted computational metric based on the computational metric comprises inputting the computational metric into a learning slope progression learning model.
7. The system of claim 1 , wherein the instructions further cause the one or more processors to, based on at least one of a computational metric trend and a predicted computational metric of a past task model, allocate a computing resource for computing of the machine learning task by the current task model prior to runtime of the current task model.
8. A method comprising:
computing, by a computing system, a machine learning task by executing an additional copy of a current task model, wherein parameters of the additional copy of the current task model are learned through training of the additional copy of the current task model;
quantifying, by the computing system, a computational metric of the additional copy of the current task model during runtime of the additional copy of the current task model; and
changing, by the computing system, allocation of a computing resource for computing of the machine learning task by the current task model during runtime of the additional copy of the current task model.
9. The method of claim 8 , wherein the computational metric comprises a sparsity measure.
10. The method of claim 8 , wherein the computational metric comprises a learning slope.
11. The method of claim 8 , further comprising determining, by the computing system, a computational metric trend based on the computational metric.
12. The method of claim 8 , further comprising deriving, by the computing system, a predicted computational metric of the additional copy of the current task model based on the computational metric.
13. The method of claim 12 , wherein deriving a predicted computational metric based on the computational metric comprises inputting, by the computing system, the computational metric into a learning slope progression learning model.
14. The method of claim 8 , further comprising the computing system, based on at least one of a computational metric trend and a predicted computational metric of a past task model, allocating a computing resource for computing of the machine learning task by the current task model prior to runtime of the current task model.
15. A massive computing system comprising:
a plurality of nodes and a plurality of edge nodes, each of the plurality of nodes and plurality of edge nodes comprising:
one or more processors; and
one or more non-transitory computer-readable media storing computer-executable instructions; that, when executed by the one or more processors, cause the one or more processors to:
compute a machine learning task by executing an additional copy of a current task model, wherein parameters of the additional copy of the current task model are learned through training of the additional copy of the current task model;
quantify a computational metric of the additional copy of the current task model during runtime of the additional copy of the current task model; and
change allocation of a computing resource for computing of the machine learning task by the current task model during runtime of the additional copy of the current task model.
16. The massive computing system of claim 15 , wherein the computational metric comprises a sparsity measure.
17. The massive computing system of claim 15 , wherein the computational metric comprises a learning slope.
18. The massive computing system of claim 15 , wherein the instructions further cause the one or more processors to determine a computational metric trend based on the computational metric.
19. The massive computing system of claim 15 , wherein the instructions further cause the one or more processors to derive a predicted computational metric of the additional copy of the current task model based on the computational metric.
20. The massive computing system of claim 19 , wherein deriving a predicted computational metric based on the computational metric comprises inputting the computational metric into a learning slope progression learning model.