gputraining

Documentation

Overview

Package gputraining defines GPU training process.

Index

• type DataParallelismMultiGPUTrainer
  • func (t DataParallelismMultiGPUTrainer) Train()

Types

type DataParallelismMultiGPUTrainer

type DataParallelismMultiGPUTrainer struct {
	TensorOperators  []*gputensor.GPUOperator
	Networks         []training.Network
	DataSource       []training.DataSource
	LossFunc         []training.LossFunction
	OptimizationAlg  []optimization.Alg
	Tester           []*training.Tester
	Epoch            int
	MaxBatchPerEpoch int
	BatchSize        int
	ShowBatchInfo    bool
	GPUs             []int
	Contexts         []*driver.Context
	Driver           *driver.Driver
}

DataParallelismMultiGPUTrainer can use multiple GPUs to train the DNN model in the data parallelism style.

func (DataParallelismMultiGPUTrainer) Train

func (t DataParallelismMultiGPUTrainer) Train()

Train will run the training algorithm on the network.