Optimizer based prunner for neural networks
AssigneeSnap Inc.
InventorsEnxu Yan, Sergey Tulyakov, Aleksei Podkin, Aleksei Stoliar
Life of this patent
Estimating the term…
Abstract
A neural network pruning system can sparsely prune neural network models using an optimizer based approach that is agnostic to the model architecture being pruned. The neural network pruning system can prune by operating on the parameter vector of the full model and the gradient vector of the loss function with respect to the model parameters. The neural network pruning system can iteratively update parameters based on the gradients, while zeroing out as many parameters as possible based a preconfigured penalty.42188
Claims20
A method comprising:
- training, at a server, a full generative neural network model;
- generating a sparsely pruned generative neural network model from the full generative neural network model;
- and
- communicating the sparsely pruned generative neural network model to a client device, the client device configured to generate a modified image based on the sparsely pruned generative neural network model.
A server comprising:
- one or more processors of the server;
- and
- a memory storing instructions that, when executed by the one or more processors, cause the server to perform operations comprising:
- training, at the server, a full generative neural network model;
- generating a sparsely pruned generative neural network model from the full generative neural network model;
- and
- communicating the sparsely pruned generative neural network model to a client device, the client device configured to generate a modified image based on the sparsely pruned generative neural network model.
A non-transitory machine-readable storage device embodying instructions that, when executed by a server, cause the server to perform operations comprising:
- training, at the server, a full generative neural network model;
- generating a sparsely pruned generative neural network model from the full generative neural network model;
- and
- communicating the sparsely pruned generative neural network model to a client device, the client device configured to generate a modified image based on the sparsely pruned generative neural network model.
Dependencies are inferred from the claim text.