train

Sep 21, 2025

Changes in this version

+ type HasShortName interface

+ ShortName func() string

Jul 1, 2025

Changes in this version

type Trainer

+ func (r *Trainer) GlobalStep() int64

+ func (r *Trainer) NumAccumulatingSteps() int

+ func (r *Trainer) WithMaxExecutors(maxExecutors int) *Trainer

Jun 12, 2025

Changes in this version

+ const AccumulatedGradientsScope

+ const TrainerLossNoRegularizationGraphParamKey

+ func GetLossNoRegularization(ctx *context.Context, g *graph.Graph) (loss *graph.Node)

+ func SetLossNoRegularization(ctx *context.Context, loss *graph.Node)

+ type OptimizeWithGradients interface

+ UpdateGraphWithGradients func(ctx *context.Context, gradients []*graph.Node, lossDType dtypes.DType)

type Trainer

+ func (r *Trainer) AccumulateGradients(numAccumulatingSteps int) error

Feb 26, 2025

Changes in this version

+ func ExecPerStepUpdateGraphFn(ctx *context.Context, g *graph.Graph)

Oct 24, 2024

Changes in this version

+ type DatasetCustomOwnership interface

+ IsOwnershipTransferred func() bool

Oct 7, 2024

Changes in this version

+ const TrainLastStepVarName

+ func GetTrainLastStepVar(ctx *context.Context) *context.Variable

type Trainer

+ func (r *Trainer) ResetComputationGraphs()

Aug 16, 2024

Changes in this version

+ const BatchNormalizationUpdatePhase

type GraphType

+ const BatchNormAveragesType

type Trainer

+ func (r *Trainer) BatchNormalizationAveragesUpdate(ds Dataset)

Jun 26, 2023

Changes in this version

+ type GraphType int

+ const EvalType

+ const TrainType

+ type OnExecFn func(exec *context.Exec, graphType GraphType)

type Trainer

+ func (r *Trainer) OnExecCreation(handler OnExecFn)

Jun 16, 2023

Changes in this version

+ func ExponentialCallback(loop *Loop, startStep int, exponentialFactor float64, callOnEnd bool, ...)

+ func PeriodicCallback(loop *Loop, period time.Duration, callOnEnd bool, name string, ...)

type Loop

+ TrainStepDurations []time.Duration

+ func (loop *Loop) MedianTrainStepDuration() time.Duration

+ func (loop *Loop) ReadGlobalStep(ctx *context.Context)

May 20, 2023

Changes in this version

type Loop

+ func (loop *Loop) FreeInputs() *Loop

Apr 23, 2023

Changes in this version

+ const TrainerAbsoluteScope

+ const TrainerLossGraphParamKey

+ const TrainerPerStepUpdateGraphFnParamKey

+ var DefaultMaxExecutors = 20

+ func AddLoss(ctx *context.Context, loss *graph.Node)

+ func AddPerStepUpdateGraphFn(ctx *context.Context, g *graph.Graph, fn ContextGraphFn)

+ func EveryNSteps(loop *Loop, n int, name string, priority Priority, fn OnStepFn)

+ func GetLosses(ctx *context.Context, g *graph.Graph) (loss *graph.Node)

+ func NTimesDuringLoop(loop *Loop, n int, name string, priority Priority, fn OnStepFn)

+ type ContextGraphFn func(ctx *context.Context, g *graph.Graph)

+ type Dataset interface

+ Name func() string

+ Reset func()

+ Yield func() (spec any, inputs []tensor.Tensor, labels []tensor.Tensor, err error)

+ type Loop struct

+ EndStep int

+ Epoch int

+ LoopStep int

+ SharedData map[string]any

+ StartStep int

+ Trainer *Trainer

+ func NewLoop(trainer *Trainer) *Loop

+ func (loop *Loop) OnEnd(name string, priority Priority, fn OnEndFn)

+ func (loop *Loop) OnStart(name string, priority Priority, fn OnStartFn)

+ func (loop *Loop) OnStep(name string, priority Priority, fn OnStepFn)

+ func (loop *Loop) RunEpochs(ds Dataset, epochs int) (metrics []tensor.Tensor, err error)

+ func (loop *Loop) RunSteps(ds Dataset, steps int) (metrics []tensor.Tensor, err error)

+ type LossFn func(labels, predictions []*graph.Node) *graph.Node

+ type ModelFn func(ctx *context.Context, spec any, inputs []*graph.Node) (predictions []*graph.Node)

+ type OnEndFn func(loop *Loop, metrics []tensor.Tensor) error

+ type OnStartFn func(loop *Loop, ds Dataset) error

+ type OnStepFn func(loop *Loop, metrics []tensor.Tensor) error

+ type Priority int

+ type Trainer struct

+ func NewTrainer(manager *graph.Manager, ctx *context.Context, modelFn ModelFn, lossFn LossFn, ...) *Trainer

+ func (r *Trainer) Context() *context.Context

+ func (r *Trainer) Eval(ds Dataset) (lossAndMetrics []tensor.Tensor, err error)

+ func (r *Trainer) EvalMetrics() []metrics.Interface

+ func (r *Trainer) EvalStep(spec any, inputs, labels []tensor.Tensor) (metrics []tensor.Tensor, err error)

+ func (r *Trainer) GetEvalExec(spec any) *context.Exec

+ func (r *Trainer) GetTrainExec(spec any) *context.Exec

+ func (r *Trainer) InDevice(deviceNum int) *Trainer

+ func (r *Trainer) Metrics() []metrics.Interface

+ func (r *Trainer) ResetTrainMetrics() error

+ func (r *Trainer) SetContext(ctx *context.Context) *Trainer

+ func (r *Trainer) TrainMetrics() []metrics.Interface

+ func (r *Trainer) TrainStep(spec any, inputs, labels []tensor.Tensor) (metrics []tensor.Tensor, err error)