An interface for classes that perform optimization on Nets - this class serves as the base class for all solvers. More...

Inheritance diagram for MyCaffe.solvers.Solver< T >:

Public Member Functions
	Solver (CudaDnn< T > cuda, Log log, SolverParameter p, CancelEvent evtCancel, AutoResetEvent evtForceSnapshot, AutoResetEvent evtForceTest, IXDatabaseBase db, IXPersist< T > persist, int nSolverCount=1, int nSolverRank=0, Net< T > shareNet=null, onGetWorkspace getws=null, onSetWorkspace setws=null)
	The Solver constructor. More...

void	Dispose ()
	Discards the resources (GPU and Host) used by this Solver. More...

bool	ForceOnTrainingIterationEvent ()
	Force an OnTrainingIterationEvent to fire. More...

void	Init (SolverParameter p, Net< T > shareNet=null)
	Initializes the Solver. More...

void	Reset ()
	Reset the iterations of the net. More...

virtual void	Solve (int nIterationOverride=-1, byte[] rgWeights=null, byte[] rgState=null, TRAIN_STEP step=TRAIN_STEP.NONE)
	The main entry of the solver function. In default, iter will be zero. Pass in a non-zero iter number to resume training for a pre-trained net. More...

bool	Step (int nIters, TRAIN_STEP step=TRAIN_STEP.NONE, bool bZeroDiffs=true, bool bApplyUpdates=true, bool bDisableOutput=false, bool bDisableProgress=false, double? dfLossOverride=null, bool? bAllowSnapshot=null)
	Steps a set of iterations through a training cycle. More...

void	Restore (byte[] rgWeights, byte[] rgState, string strSkipBlobTypes=null)
	The restore method simply calls the RestoreSolverState method of the inherited class. More...

void	Snapshot (bool bForced, bool bScheduled, bool bUpdateDatabase=true)
	The snapshot function implements the basic snapshotting utility that stores the learned net. This method calls the SnapshotSolverState method of the inherited class. More...

SnapshotArgs	GetSnapshotArgs (byte[] rgState, byte[] rgWeights, double dfAccuracy, double dfError, int nIteration, SNAPSHOT_WEIGHT_UPDATE_METHOD wtUpdt)
	The GetSnapshotArgs method fills out a snapshot args structure. More...

double	TestAll (int nIterationOverride=-1)
	Run a TestAll by running all test Nets. More...

double	TestDetection (int nIterationOverride=-1, int nTestNetId=0)
	Run an SSD detection test on a given test Net by running it through its iterations. More...

double	TestClassification (int nIterationOverride=-1, int nTestNetId=0)
	Run a test on a given test Net by running it through its iterations. More...

void	UpdateSmoothedLoss (double dfLoss, int nStartIter, int nAverageLoss=0)
	Update the avaraged loss value. More...

abstract double	ApplyUpdate (int nIterationOverride=-1)
	Make and apply the update value for the current iteration. More...

Static Public Member Functions
static SGDSolver< T >	Create (CudaDnn< T > cuda, Log log, ProjectEx p, CancelEvent evtCancel, AutoResetEvent evtForceSnapshot, AutoResetEvent evtForceTest, IXDatabaseBase db, IXPersist< T > persist, int nSolverCount=1, int nSolverRank=0, Net< T > shareNet=null, onGetWorkspace getws=null, onSetWorkspace setws=null)
	Create a new Solver based on the project containing the SolverParameter. More...

static SGDSolver< T >	Create (CudaDnn< T > cuda, Log log, SolverParameter solverParam, CancelEvent evtCancel, AutoResetEvent evtForceSnapshot, AutoResetEvent evtForceTest, IXDatabaseBase db, IXPersist< T > persist, int nSolverCount=1, int nSolverRank=0, Net< T > shareNet=null, onGetWorkspace getws=null, onSetWorkspace setws=null)
	Create a new Solver based on the project containing the SolverParameter. More...

Protected Member Functions
virtual void	dispose ()
	Override that allows discarding of resources (GPU and Host) used by this Solver. More...

void	InitTrainNet (Net< T > shareNet=null)
	Initializes the Net used by the solver for training. More...

void	InitTestNets ()
	Initializes the Net used by the Solver for testing. More...

abstract byte[]	SnapshotSolverState ()
	Save the current solver state. More...

abstract void	RestoreSolverState (byte[] rgState)
	Restore a solver state. More...

Protected Attributes
CudaDnn< T >	m_cuda
	Specifies the instance of CudaDnn used by the Solver that provides a connection to Cuda. More...

Log	m_log
	Specifies the Log for output. More...

SolverParameter	m_param
	Specifies the SolverParameter that defines how the Solver operates. More...

Net< T >	m_net
	Specifies the training Net. More...

List< Net< T > >	m_rgTestNets = new List<Net<T>>()
	Specifies the testing Nets. More...

int	m_nIter
	Specifies the current iteration. More...

int	m_nCurrentStep
	Specifies the current step. More...

List< double >	m_rgLosses = new List<double>()
	Specifies the Losses used to calculate the smoothed Loss. More...

double	m_dfSmoothedLoss = 0
	Specifies the smoothed loss protected for derived classes to use. More...

double?	m_dfIterAccuracy = null
	Specifies the iteration accuracy calculated when a blob exists with the name 'accuracy'. More...

int	m_nSolverCount = 1
	Specifies the Solver count in a multi-GPU training session. More...

int	m_nSolverRank = 0
	Specifies the Solver rank of this solver, where rank == 0 is the root Solver. More...

IXPersist< T >	m_persist
	Specifies the persistance object used to save weight and solver states. More...

double	m_dfLearningRateOverride = 0
	Optionally, specifies a learning rate override (default = 0, which ignores this setting). More...

Properties
double	LearningRateOverride `[getset]`
	Get/set the learning rate override. When 0, this setting is ignored. More...

int	TrainingTimeLimitInMinutes `[getset]`
	Get/set the training time limit in minutes. When set to 0, no time limit is imposed on training. More...

SNAPSHOT_WEIGHT_UPDATE_METHOD	SnapshotWeightUpdateMethod `[getset]`
	Get/set the snapshot weight update method. More...

IXDatabaseBase	Database `[get]`
	Returns the in-memory MyCaffeDatabase used. More...

bool	EnableTesting `[getset]`
	When enabled, the training cycle calls TestAll periodically based on the SolverParameter. Otherwise testing is not performed. More...

bool	EnableBlobDebugging `[getset]`
	When enabled, the OnTrainingIteration event is set extra debugging information describing the state of each Blob used by the Solver. More...

bool	EnableLayerDebugging `[getset]`
	Enable/disable layer debugging which causes each layer to check for NAN/INF on each forward/backward pass and throw an exception when found. More...

bool	EnableBreakOnFirstNaN `[getset]`
	When enabled (requires EnableBlobDebugging = true), the Solver immediately stop training upon detecting the first NaN withing the training Net. More...

bool	EnableDetailedNanDetection `[getset]`
	When enabled (requires EnableBlobDebugging = true), the detailed Nan (and Infinity) detection is perofmed on each blob when training Net. More...

bool	EnableSingleStep `[getset]`
	When enabled (requires EnableBlobDebugging = true), the Solver only runs one training cycle. More...

bool	WeightsUpdated `[getset]`
	Get/set when the weights have been updated. More...

object	Tag `[getset]`
	Returns a generic tag associated with the Solver. More...

Net< T >	TestingNet `[get]`
	Returns the testing Net used by the solver. More...

Net< T >	TrainingNet `[get]`
	Returns the training Net used by the solver. More...

CudaDnn< T >	Cuda `[get]`
	Returns the CudaDnn instance used by the Solver. More...

string	ActiveLabelCounts `[get]`
	Returns a string describing the labels detected in the training along with the % that each label has participated in the training. More...

string	LabelQueryHitPercents `[get]`
	Return the label query hit percentages for the active datasource. More...

string	LabelQueryEpochs `[get]`
	Return the label query epochs for the active datasource. More...

int	CurrentIteration `[get]`
	Returns the current training iteration. More...

int	MaximumIteration `[get]`
	Returns the maximum training iterations. More...

int	TrainingIterations `[get]`
	Returns the current training iterations remaining. More...

int?	TestingIterations `[get]`
	Returns the current testing iterations remaining. More...

int	TrainingIterationOverride `[getset]`
	Get/set the training iteration override. More...

int	TestingIterationOverride `[getset]`
	Get/set the testing iteration override. More...

AutoResetEvent	CompletedEvent `[get]`
	Returns an auto reset event that is set upon training completion. More...

CancelEvent	CancelEvent `[get]`
	Returns the cancel event which when set cancels the current operation run by the Solver. More...

double	smoothed_loss `[get]`
	Returns the smoothed loss. More...

SolverParameter	parameter `[get]`
	Returns the SolverParameter used. More...

Net< T >	net `[get]`
	Returns the main training Net. More...

List< Net< T > >	test_nets `[get]`
	Returns the testing Nets. More...

int	iter `[get]`
	Returns the current training iteration. More...

SolverParameter.SolverType	type `[get]`
	Returns the type of solver. More...

bool	forceSnapshot `[get]`
	Returns whether or not a snapshot has been forced. More...

bool	forceTest `[get]`
	Returns whether or not a test has been forced. More...

int	solver_count `[get]`
	Returns the solver count in a multi-GPU session. More...

int	solver_rank `[get]`
	Returns this Solver's rank in a multi-GPU session. More...

bool?	is_root_solver `[get]`
	Returns whether or not this is the root solver. More...

Events
EventHandler	OnStart
	The OnStart event fires at the start of each training iteration. More...

EventHandler	OnAborted
	The OnAborted event fires after aborting a training cycle. More...

EventHandler< GradientsReadyArgs >	OnGradientsReady
	The OnGradientsReady event fires after the gradients of a Solver are ready for distribution to other Solvers in a multi-GPU training session. More...

EventHandler< SnapshotArgs >	OnSnapshot
	The OnSnapshot event fires when the Solver detects that a snapshot is needed. More...

EventHandler< TrainingIterationArgs< T > >	OnTrainingIteration
	The OnTrainingIteration event fires at the end of each training iteration. More...

EventHandler< TestingIterationArgs< T > >	OnTestingIteration
	The OnTestingIteration event fires at the end of each testing iteration. More...

EventHandler< TestResultArgs< T > >	OnTestResults
	When specified, the OnTestResults event fires after each single test run. The recipient is responsible for setting the accuracy. More...

EventHandler< TestArgs >	OnTest
	When specified, the OnTest event fires during a TestAll and overrides the call to Test. More...

EventHandler	OnTestStart
	The OnTestStart event fires at the start of each testing iteration. More...

EventHandler< CustomForwardBackArgs< T > >	OnCustomForwardBack
	The OnCustomForwardBack allows for overriding the forward/backward operations within the solver. More...

EventHandler< WorkspaceArgs >	OnGetWorkspace
	Specifies the OnGetWorkspace event that fires when the getWorkspace() function is called by a layer to get a shareable workspace to conserve GPU memory. More...

EventHandler< WorkspaceArgs >	OnSetWorkspace
	Specifies the OnSetWorkspace event that fires when the setWorkspace() function is called by a layer to get a shareable workspace to conserve GPU memory. More...

Detailed Description

An interface for classes that perform optimization on Nets - this class serves as the base class for all solvers.

Requires implementation of ApplyUpdate to compute a parameter update given the current state of the Net parameters.

Template Parameters

T	Specifies the base type float or double. Using float is recommended to conserve GPU memory.

Definition at line 27 of file Solver.cs.

Constructor & Destructor Documentation

◆ Solver()

MyCaffe.solvers.Solver< T >.Solver	(	CudaDnn< T >	cuda,
		Log	log,
		SolverParameter	p,
		CancelEvent	evtCancel,
		AutoResetEvent	evtForceSnapshot,
		AutoResetEvent	evtForceTest,
		IXDatabaseBase	db,
		IXPersist< T >	persist,
		int	nSolverCount = `1`,
		int	nSolverRank = `0`,
		Net< T >	shareNet = `null`,
		onGetWorkspace	getws = `null`,
		onSetWorkspace	setws = `null`
	)

The Solver constructor.

Parameters

cuda	Specifies the instance of CudaDnn to use.
log	Specifies the Log for output.
p	Specifies teh SolverParameter.
evtCancel	Specifies a CancelEvent used to cancel the current operation (e.g. training, testing) for which the Solver is performing.
evtForceSnapshot	Specifies an automatic reset event that causes the Solver to perform a Snapshot when set.
evtForceTest	Specifies an automatic reset event that causes teh Solver to run a testing cycle when set.
db	Specifies the in-memory MyCaffeDatabase.
persist	Specifies the peristence used for loading and saving weights.
nSolverCount	Specifies the number of Solvers participating in a multi-GPU session.
nSolverRank	Specifies the rank of this Solver in a multi-GPU session.
shareNet	Optionally, specifies the net to share when creating the training network (default = null, meaning no share net is used).
getws	Optionally, specifies the handler for getting the workspace.
setws	Optionally, specifies the handler for setting the workspace.

Definition at line 181 of file Solver.cs.

Member Function Documentation

◆ ApplyUpdate()

abstract double MyCaffe.solvers.Solver< T >.ApplyUpdate ( int nIterationOverride = -1 )

pure virtual

Make and apply the update value for the current iteration.

Parameters

nIterationOverride Optionally, specifies an iteration override, or -1 which is ignored.

Implemented in MyCaffe.solvers.LBFGSSolver< T >, and MyCaffe.solvers.SGDSolver< T >.

◆ Create() [1/2]

static SGDSolver< T > MyCaffe.solvers.Solver< T >.Create	(	CudaDnn< T >	cuda,
		Log	log,
		ProjectEx	p,
		CancelEvent	evtCancel,
		AutoResetEvent	evtForceSnapshot,
		AutoResetEvent	evtForceTest,
		IXDatabaseBase	db,
		IXPersist< T >	persist,
		int	nSolverCount = `1`,
		int	nSolverRank = `0`,
		Net< T >	shareNet = `null`,
		onGetWorkspace	getws = `null`,
		onSetWorkspace	setws = `null`
	)

static

Create a new Solver based on the project containing the SolverParameter.

Parameters

cuda	Specifies the CudaDnn instance that the new Solver will use.
log	Specifies the Log for output that the new Solver will use.
p	Specifies the project to use in initialzing the new Solver.
evtCancel	Specifies the CancelEvent that the new Solver will use.
evtForceSnapshot	Specifies the force snapshot event that the new Solver will use.
evtForceTest	Specifies the force test event that the new Solver will use.
db	Specifies the in-memory MyCaffeDatabase that the new Solver will use.
persist	Specifies the peristence used for loading and saving weights.
nSolverCount	Specifies the number of Solvers participating in a multi-GPu session.
nSolverRank	Specifies the rank of the new Solver.
shareNet	Optionally, specifies the net to share when creating the training network (default = null, meaning no share net is used).
getws	Optionally, specifies the handler for getting the workspace.
setws	Optionally, specifies the handler for setting the workspace.

Returns: A new Solver instance is returned.

Definition at line 1889 of file Solver.cs.

◆ Create() [2/2]

static SGDSolver< T > MyCaffe.solvers.Solver< T >.Create	(	CudaDnn< T >	cuda,
		Log	log,
		SolverParameter	solverParam,
		CancelEvent	evtCancel,
		AutoResetEvent	evtForceSnapshot,
		AutoResetEvent	evtForceTest,
		IXDatabaseBase	db,
		IXPersist< T >	persist,
		int	nSolverCount = `1`,
		int	nSolverRank = `0`,
		Net< T >	shareNet = `null`,
		onGetWorkspace	getws = `null`,
		onSetWorkspace	setws = `null`
	)

static

Create a new Solver based on the project containing the SolverParameter.

Parameters

cuda	Specifies the CudaDnn instance that the new Solver will use.
log	Specifies the Log for output that the new Solver will use.
solverParam	Specifies the SolverParameter used to create the Solver.
evtCancel	Specifies the CancelEvent that the new Solver will use.
evtForceSnapshot	Specifies the force snapshot event that the new Solver will use.
evtForceTest	Specifies the force test event that the new Solver will use.
db	Specifies the in-memory MyCaffeDatabase that the new Solver will use.
persist	Specifies the peristence used for loading and saving weights.
nSolverCount	Specifies the number of Solvers participating in a multi-GPu session.
nSolverRank	Specifies the rank of the new Solver.
shareNet	Optionally, specifies the net to share when creating the training network (default = null, meaning no share net is used).
getws	Optionally, specifies the handler for getting the workspace.
setws	Optionally, specifies the handler for setting the workspace.

Returns

Definition at line 1930 of file Solver.cs.

◆ Dispose()

void MyCaffe.solvers.Solver< T >.Dispose ( )

Discards the resources (GPU and Host) used by this Solver.

Definition at line 218 of file Solver.cs.

◆ dispose()

virtual void MyCaffe.solvers.Solver< T >.dispose ( )

protectedvirtual

Override that allows discarding of resources (GPU and Host) used by this Solver.

Reimplemented in MyCaffe.solvers.LBFGSSolver< T >, and MyCaffe.solvers.SGDSolver< T >.

Definition at line 317 of file Solver.cs.

◆ ForceOnTrainingIterationEvent()

bool MyCaffe.solvers.Solver< T >.ForceOnTrainingIterationEvent ( )

Force an OnTrainingIterationEvent to fire.

Returns: When fired, true is returned, otherwise false.

Definition at line 236 of file Solver.cs.

◆ GetSnapshotArgs()

SnapshotArgs MyCaffe.solvers.Solver< T >.GetSnapshotArgs	(	byte[]	rgState,
		byte[]	rgWeights,
		double	dfAccuracy,
		double	dfError,
		int	nIteration,
		SNAPSHOT_WEIGHT_UPDATE_METHOD	wtUpdt
	)

The GetSnapshotArgs method fills out a snapshot args structure.

Parameters

rgState	Specifies the state bytes or null.
rgWeights	Specifies the weight bytes or null.
dfAccuracy	Specifies the accuracy.
dfError	Specifies the error.
nIteration	Specifies the interation.
wtUpdt	Specifies the weight update method.

Returns: The args are returned.

Definition at line 1159 of file Solver.cs.

◆ Init()

void MyCaffe.solvers.Solver< T >.Init	(	SolverParameter	p,
		Net< T >	shareNet = `null`
	)

Initializes the Solver.

Parameters

p	Specifies the SolverParameters used to initialize the Solver.
shareNet	Optionally, specifies a network to share (default = null).

Definition at line 454 of file Solver.cs.

◆ InitTestNets()

void MyCaffe.solvers.Solver< T >.InitTestNets ( )

protected

Initializes the Net used by the Solver for testing.

Definition at line 579 of file Solver.cs.

◆ InitTrainNet()

void MyCaffe.solvers.Solver< T >.InitTrainNet ( Net< T > shareNet = null )

protected

Initializes the Net used by the solver for training.

Parameters

shareNet Optionally, specifies a network to share (default = null).

Definition at line 488 of file Solver.cs.

◆ Reset()

void MyCaffe.solvers.Solver< T >.Reset ( )

Reset the iterations of the net.

Definition at line 478 of file Solver.cs.

◆ Restore()

void MyCaffe.solvers.Solver< T >.Restore	(	byte[]	rgWeights,
		byte[]	rgState,
		string	strSkipBlobTypes = `null`
	)

The restore method simply calls the RestoreSolverState method of the inherited class.

Parameters

rgWeights	Specifies the weights to load, or null to ignore.
rgState	Specifies the state to load, or null to ignore.
strSkipBlobTypes	Specifies the blob types to ignore and not load, or null to ignore.

Definition at line 1097 of file Solver.cs.

◆ RestoreSolverState()

abstract void MyCaffe.solvers.Solver< T >.RestoreSolverState ( byte[] rgState )

protectedpure virtual

Restore a solver state.

Implemented in MyCaffe.solvers.LBFGSSolver< T >, and MyCaffe.solvers.SGDSolver< T >.

◆ Snapshot()

void MyCaffe.solvers.Solver< T >.Snapshot	(	bool	bForced,
		bool	bScheduled,
		bool	bUpdateDatabase = `true`
	)

The snapshot function implements the basic snapshotting utility that stores the learned net. This method calls the SnapshotSolverState method of the inherited class.

Parameters

bForced	Specifies whehter or not to force the snapshot.
bScheduled	Specifies whether or not the snapshot is a scheduled snapshot that occurs at regular intervals, or a snapshot based on an improving accuracy.
bUpdateDatabase	Optionally, specifies to update the database (default = true).

Definition at line 1115 of file Solver.cs.

◆ SnapshotSolverState()

abstract byte[] MyCaffe.solvers.Solver< T >.SnapshotSolverState ( )

protectedpure virtual

Save the current solver state.

Implemented in MyCaffe.solvers.LBFGSSolver< T >, and MyCaffe.solvers.SGDSolver< T >.

◆ Solve()

virtual void MyCaffe.solvers.Solver< T >.Solve	(	int	nIterationOverride = `-1`,
		byte[]	rgWeights = `null`,
		byte[]	rgState = `null`,
		TRAIN_STEP	step = `TRAIN_STEP.NONE`
	)

virtual

The main entry of the solver function. In default, iter will be zero. Pass in a non-zero iter number to resume training for a pre-trained net.

Parameters

nIterationOverride	Optionally, specifies an iteration override value to use for the number of iterations run. The default is -1, which ignores the parameter.
rgWeights	Optionally, specifies weights to load via the Restore method. The default is null which ignores the parameter.
rgState	Optionally, specifies the state to load via the Restore method. The default is null which ignores the parameter.
step	Optionally, specifies to single step the training pass - typically this is used during debugging. The default = TRAIN_STEP.NONE which runs the solver in the normal manner.

Definition at line 744 of file Solver.cs.

◆ Step()

bool MyCaffe.solvers.Solver< T >.Step	(	int	nIters,
		TRAIN_STEP	step = `TRAIN_STEP.NONE`,
		bool	bZeroDiffs = `true`,
		bool	bApplyUpdates = `true`,
		bool	bDisableOutput = `false`,
		bool	bDisableProgress = `false`,
		double?	dfLossOverride = `null`,
		bool?	bAllowSnapshot = `null`
	)

Steps a set of iterations through a training cycle.

Parameters

nIters	Specifies the number of steps to iterate.
step	Optionally, specifies to single step the training pass - typically this is used during debugging. The default = TRAIN_STEP.NONE for no stepping.
bZeroDiffs	Optionally, specifies whether or not to zero out the gradient diffs (default = true).
bApplyUpdates	Optionally, specifies to apply the gradient updates to the weights (default = true).
bDisableOutput	Optionally, disable the output to the log.
bDisableProgress	Optionally, disables the progress updating to the log.
dfLossOverride	Optionally, specifies a loss override which can be useful when using a backward step only.
bAllowSnapshot	Optionally, specifies whether or not a snapshot is allowed even during TRAIN_STEP.

Returns

Definition at line 818 of file Solver.cs.

◆ TestAll()

double MyCaffe.solvers.Solver< T >.TestAll ( int nIterationOverride = -1 )

Run a TestAll by running all test Nets.

Depending on the eval_type tests are run as Test (default), TestClassification (for SSD Classification), or TestDetection (for SSD Detection).

Parameters

nIterationOverride Specifies an override to the iterations to run.

Returns: The accuracy of the testing run is returned as a percentage value in the range [0, 1].

Definition at line 1322 of file Solver.cs.

◆ TestClassification()

double MyCaffe.solvers.Solver< T >.TestClassification	(	int	nIterationOverride = `-1`,
		int	nTestNetId = `0`
	)

Run a test on a given test Net by running it through its iterations.

Parameters

nIterationOverride	Specifies an override the the number of iterations to run.
nTestNetId	Specifies the ID of the test Net to run.

Returns: The accuracy of the test run is returned as a percentage in the range [0, 1].

Definition at line 1597 of file Solver.cs.

◆ TestDetection()

double MyCaffe.solvers.Solver< T >.TestDetection	(	int	nIterationOverride = `-1`,
		int	nTestNetId = `0`
	)

Run an SSD detection test on a given test Net by running it through its iterations.

Parameters

nIterationOverride	Specifies an override the the number of iterations to run.
nTestNetId	Specifies the ID of the test Net to run.

Returns: The accuracy of the test run is returned as a percentage in the range [0, 1].

Definition at line 1396 of file Solver.cs.

◆ UpdateSmoothedLoss()

void MyCaffe.solvers.Solver< T >.UpdateSmoothedLoss	(	double	dfLoss,
		int	nStartIter,
		int	nAverageLoss = `0`
	)

Update the avaraged loss value.

Parameters

dfLoss	Specifies the new loss value to add into the average.
nStartIter	Specifies the starting iteration.
nAverageLoss	Optionally, specifies the number of iterations to average over (default = param.average_loss).

Definition at line 1826 of file Solver.cs.

Member Data Documentation

◆ m_cuda

CudaDnn<T> MyCaffe.solvers.Solver< T >.m_cuda

protected

Specifies the instance of CudaDnn used by the Solver that provides a connection to Cuda.

Definition at line 32 of file Solver.cs.

◆ m_dfIterAccuracy

double? MyCaffe.solvers.Solver< T >.m_dfIterAccuracy = null

protected

Specifies the iteration accuracy calculated when a blob exists with the name 'accuracy'.

Definition at line 74 of file Solver.cs.

◆ m_dfLearningRateOverride

double MyCaffe.solvers.Solver< T >.m_dfLearningRateOverride = 0

protected

Optionally, specifies a learning rate override (default = 0, which ignores this setting).

Definition at line 94 of file Solver.cs.

◆ m_dfSmoothedLoss

double MyCaffe.solvers.Solver< T >.m_dfSmoothedLoss = 0

protected

Specifies the smoothed loss protected for derived classes to use.

Definition at line 70 of file Solver.cs.

◆ m_log

Log MyCaffe.solvers.Solver< T >.m_log

protected

Specifies the Log for output.

Definition at line 36 of file Solver.cs.

◆ m_nCurrentStep

int MyCaffe.solvers.Solver< T >.m_nCurrentStep

protected

Specifies the current step.

Definition at line 56 of file Solver.cs.

◆ m_net

Net<T> MyCaffe.solvers.Solver< T >.m_net

protected

Specifies the training Net.

Definition at line 44 of file Solver.cs.

◆ m_nIter

int MyCaffe.solvers.Solver< T >.m_nIter

protected

Specifies the current iteration.

Definition at line 52 of file Solver.cs.

◆ m_nSolverCount

int MyCaffe.solvers.Solver< T >.m_nSolverCount = 1

protected

Specifies the Solver count in a multi-GPU training session.

Definition at line 82 of file Solver.cs.

◆ m_nSolverRank

int MyCaffe.solvers.Solver< T >.m_nSolverRank = 0

protected

Specifies the Solver rank of this solver, where rank == 0 is the root Solver.

Definition at line 86 of file Solver.cs.

◆ m_param

SolverParameter MyCaffe.solvers.Solver< T >.m_param

protected

Specifies the SolverParameter that defines how the Solver operates.

Definition at line 40 of file Solver.cs.

◆ m_persist

IXPersist<T> MyCaffe.solvers.Solver< T >.m_persist

protected

Specifies the persistance object used to save weight and solver states.

Definition at line 90 of file Solver.cs.

◆ m_rgLosses

List<double> MyCaffe.solvers.Solver< T >.m_rgLosses = new List<double>()

protected

Specifies the Losses used to calculate the smoothed Loss.

Definition at line 60 of file Solver.cs.

◆ m_rgTestNets

List<Net<T> > MyCaffe.solvers.Solver< T >.m_rgTestNets = new List<Net<T>>()

protected

Specifies the testing Nets.

Definition at line 48 of file Solver.cs.

Property Documentation

◆ ActiveLabelCounts

string MyCaffe.solvers.Solver< T >.ActiveLabelCounts

get

Returns a string describing the labels detected in the training along with the % that each label has participated in the training.

Definition at line 667 of file Solver.cs.

◆ CancelEvent

CancelEvent MyCaffe.solvers.Solver< T >.CancelEvent

get

Returns the cancel event which when set cancels the current operation run by the Solver.

Definition at line 1204 of file Solver.cs.

◆ CompletedEvent

AutoResetEvent MyCaffe.solvers.Solver< T >.CompletedEvent

get

Returns an auto reset event that is set upon training completion.

Definition at line 1196 of file Solver.cs.

◆ Cuda

CudaDnn<T> MyCaffe.solvers.Solver< T >.Cuda

get

Returns the CudaDnn instance used by the Solver.

Definition at line 659 of file Solver.cs.

◆ CurrentIteration

int MyCaffe.solvers.Solver< T >.CurrentIteration

get

Returns the current training iteration.

Definition at line 691 of file Solver.cs.

◆ Database

IXDatabaseBase MyCaffe.solvers.Solver< T >.Database

get

Returns the in-memory MyCaffeDatabase used.

Definition at line 309 of file Solver.cs.

◆ EnableBlobDebugging

bool MyCaffe.solvers.Solver< T >.EnableBlobDebugging

getset

When enabled, the OnTrainingIteration event is set extra debugging information describing the state of each Blob used by the Solver.

Definition at line 360 of file Solver.cs.

◆ EnableBreakOnFirstNaN

bool MyCaffe.solvers.Solver< T >.EnableBreakOnFirstNaN

getset

When enabled (requires EnableBlobDebugging = true), the Solver immediately stop training upon detecting the first NaN withing the training Net.

Definition at line 381 of file Solver.cs.

◆ EnableDetailedNanDetection

bool MyCaffe.solvers.Solver< T >.EnableDetailedNanDetection

getset

When enabled (requires EnableBlobDebugging = true), the detailed Nan (and Infinity) detection is perofmed on each blob when training Net.

IMPORTANT: The use of this setting is only recommended on TCC enabled drivers.

See also: NVIDIA Tesla Compute Cluster (TCC) Help

Definition at line 394 of file Solver.cs.

◆ EnableLayerDebugging

bool MyCaffe.solvers.Solver< T >.EnableLayerDebugging

getset

Enable/disable layer debugging which causes each layer to check for NAN/INF on each forward/backward pass and throw an exception when found.

This option dramatically slows down training and is only recommended during debugging.

Definition at line 372 of file Solver.cs.

◆ EnableSingleStep

bool MyCaffe.solvers.Solver< T >.EnableSingleStep

getset

When enabled (requires EnableBlobDebugging = true), the Solver only runs one training cycle.

Definition at line 403 of file Solver.cs.

◆ EnableTesting

bool MyCaffe.solvers.Solver< T >.EnableTesting

getset

When enabled, the training cycle calls TestAll periodically based on the SolverParameter. Otherwise testing is not performed.

Definition at line 351 of file Solver.cs.

◆ forceSnapshot

bool MyCaffe.solvers.Solver< T >.forceSnapshot

getprotected

Returns whether or not a snapshot has been forced.

Definition at line 1260 of file Solver.cs.

◆ forceTest

bool MyCaffe.solvers.Solver< T >.forceTest

get

Returns whether or not a test has been forced.

Definition at line 1274 of file Solver.cs.

◆ is_root_solver

bool? MyCaffe.solvers.Solver< T >.is_root_solver

get

Returns whether or not this is the root solver.

The root solver has rank = 0.

Definition at line 1308 of file Solver.cs.

◆ iter

int MyCaffe.solvers.Solver< T >.iter

get

Returns the current training iteration.

Definition at line 1244 of file Solver.cs.

◆ LabelQueryEpochs

string MyCaffe.solvers.Solver< T >.LabelQueryEpochs

get

Return the label query epochs for the active datasource.

Definition at line 683 of file Solver.cs.

◆ LabelQueryHitPercents

string MyCaffe.solvers.Solver< T >.LabelQueryHitPercents

get

Return the label query hit percentages for the active datasource.

Definition at line 675 of file Solver.cs.

◆ LearningRateOverride

double MyCaffe.solvers.Solver< T >.LearningRateOverride

getset

Get/set the learning rate override. When 0, this setting is ignored.

Definition at line 226 of file Solver.cs.

◆ MaximumIteration

int MyCaffe.solvers.Solver< T >.MaximumIteration

get

Returns the maximum training iterations.

Definition at line 699 of file Solver.cs.

◆ net

Net<T> MyCaffe.solvers.Solver< T >.net

get

Returns the main training Net.

Definition at line 1228 of file Solver.cs.

◆ parameter

SolverParameter MyCaffe.solvers.Solver< T >.parameter

get

Returns the SolverParameter used.

Definition at line 1220 of file Solver.cs.

◆ smoothed_loss

double MyCaffe.solvers.Solver< T >.smoothed_loss

get

Returns the smoothed loss.

Definition at line 1212 of file Solver.cs.

◆ SnapshotWeightUpdateMethod

SNAPSHOT_WEIGHT_UPDATE_METHOD MyCaffe.solvers.Solver< T >.SnapshotWeightUpdateMethod

getset

Get/set the snapshot weight update method.

Definition at line 300 of file Solver.cs.

◆ solver_count

int MyCaffe.solvers.Solver< T >.solver_count

get

Returns the solver count in a multi-GPU session.

Definition at line 1289 of file Solver.cs.

◆ solver_rank

int MyCaffe.solvers.Solver< T >.solver_rank

get

Returns this Solver's rank in a multi-GPU session.

Definition at line 1297 of file Solver.cs.

◆ Tag

object MyCaffe.solvers.Solver< T >.Tag

getset

Returns a generic tag associated with the Solver.

Definition at line 421 of file Solver.cs.

◆ test_nets

List<Net<T> > MyCaffe.solvers.Solver< T >.test_nets

get

Returns the testing Nets.

Definition at line 1236 of file Solver.cs.

◆ TestingIterationOverride

int MyCaffe.solvers.Solver< T >.TestingIterationOverride

getset

Get/set the testing iteration override.

Definition at line 1187 of file Solver.cs.

◆ TestingIterations

int? MyCaffe.solvers.Solver< T >.TestingIterations

get

Returns the current testing iterations remaining.

Definition at line 723 of file Solver.cs.

◆ TestingNet

Net<T> MyCaffe.solvers.Solver< T >.TestingNet

get

Returns the testing Net used by the solver.

Definition at line 430 of file Solver.cs.

◆ TrainingIterationOverride

int MyCaffe.solvers.Solver< T >.TrainingIterationOverride

getset

Get/set the training iteration override.

Definition at line 1178 of file Solver.cs.

◆ TrainingIterations

int MyCaffe.solvers.Solver< T >.TrainingIterations

get

Returns the current training iterations remaining.

Definition at line 707 of file Solver.cs.

◆ TrainingNet

Net<T> MyCaffe.solvers.Solver< T >.TrainingNet

get

Returns the training Net used by the solver.

Definition at line 444 of file Solver.cs.

◆ TrainingTimeLimitInMinutes

int MyCaffe.solvers.Solver< T >.TrainingTimeLimitInMinutes

getset

Get/set the training time limit in minutes. When set to 0, no time limit is imposed on training.

Definition at line 291 of file Solver.cs.

◆ type

SolverParameter.SolverType MyCaffe.solvers.Solver< T >.type

get

Returns the type of solver.

Definition at line 1252 of file Solver.cs.

◆ WeightsUpdated

bool MyCaffe.solvers.Solver< T >.WeightsUpdated

getset

Get/set when the weights have been updated.

Definition at line 412 of file Solver.cs.

Event Documentation

◆ OnAborted

EventHandler MyCaffe.solvers.Solver< T >.OnAborted

The OnAborted event fires after aborting a training cycle.

Definition at line 122 of file Solver.cs.

◆ OnCustomForwardBack

EventHandler<CustomForwardBackArgs<T> > MyCaffe.solvers.Solver< T >.OnCustomForwardBack

The OnCustomForwardBack allows for overriding the forward/backward operations within the solver.

Definition at line 155 of file Solver.cs.

◆ OnGetWorkspace

EventHandler<WorkspaceArgs> MyCaffe.solvers.Solver< T >.OnGetWorkspace

Specifies the OnGetWorkspace event that fires when the getWorkspace() function is called by a layer to get a shareable workspace to conserve GPU memory.

Definition at line 159 of file Solver.cs.

◆ OnGradientsReady

EventHandler<GradientsReadyArgs> MyCaffe.solvers.Solver< T >.OnGradientsReady

The OnGradientsReady event fires after the gradients of a Solver are ready for distribution to other Solvers in a multi-GPU training session.

Definition at line 126 of file Solver.cs.

◆ OnSetWorkspace

EventHandler<WorkspaceArgs> MyCaffe.solvers.Solver< T >.OnSetWorkspace

Specifies the OnSetWorkspace event that fires when the setWorkspace() function is called by a layer to get a shareable workspace to conserve GPU memory.

Definition at line 163 of file Solver.cs.

◆ OnSnapshot

EventHandler<SnapshotArgs> MyCaffe.solvers.Solver< T >.OnSnapshot

The OnSnapshot event fires when the Solver detects that a snapshot is needed.

Definition at line 130 of file Solver.cs.

◆ OnStart

EventHandler MyCaffe.solvers.Solver< T >.OnStart

The OnStart event fires at the start of each training iteration.

Definition at line 118 of file Solver.cs.

◆ OnTest

EventHandler<TestArgs> MyCaffe.solvers.Solver< T >.OnTest

When specified, the OnTest event fires during a TestAll and overrides the call to Test.

Definition at line 146 of file Solver.cs.

◆ OnTestingIteration

EventHandler<TestingIterationArgs<T> > MyCaffe.solvers.Solver< T >.OnTestingIteration

The OnTestingIteration event fires at the end of each testing iteration.

Definition at line 138 of file Solver.cs.

◆ OnTestResults

EventHandler<TestResultArgs<T> > MyCaffe.solvers.Solver< T >.OnTestResults

When specified, the OnTestResults event fires after each single test run. The recipient is responsible for setting the accuracy.

Definition at line 142 of file Solver.cs.

◆ OnTestStart

EventHandler MyCaffe.solvers.Solver< T >.OnTestStart

The OnTestStart event fires at the start of each testing iteration.

Definition at line 150 of file Solver.cs.

◆ OnTrainingIteration

EventHandler<TrainingIterationArgs<T> > MyCaffe.solvers.Solver< T >.OnTrainingIteration

The OnTrainingIteration event fires at the end of each training iteration.

Definition at line 134 of file Solver.cs.

The documentation for this class was generated from the following file:

C:/Data/Data/SS_Projects/Intelligence/GitHub/MyCaffe/MyCaffe/solvers/Solver.cs

Public Member Functions

Static Public Member Functions

Protected Member Functions

Protected Attributes

Properties

Events

Detailed Description

Constructor & Destructor Documentation

◆ Solver()

Member Function Documentation

◆ ApplyUpdate()

◆ Create() [1/2]

◆ Create() [2/2]

◆ Dispose()

◆ dispose()

◆ ForceOnTrainingIterationEvent()

◆ GetSnapshotArgs()

◆ Init()

◆ InitTestNets()

◆ InitTrainNet()

◆ Reset()

◆ Restore()

◆ RestoreSolverState()

◆ Snapshot()

◆ SnapshotSolverState()

◆ Solve()

◆ Step()

◆ TestAll()

◆ TestClassification()

◆ TestDetection()

◆ UpdateSmoothedLoss()

Member Data Documentation

◆ m_cuda

◆ m_dfIterAccuracy

◆ m_dfLearningRateOverride

◆ m_dfSmoothedLoss

◆ m_log

◆ m_nCurrentStep

◆ m_net

◆ m_nIter

◆ m_nSolverCount

◆ m_nSolverRank

◆ m_param

◆ m_persist

◆ m_rgLosses

◆ m_rgTestNets

Property Documentation

◆ ActiveLabelCounts

◆ CancelEvent

◆ CompletedEvent

◆ Cuda

◆ CurrentIteration

◆ Database

◆ EnableBlobDebugging

◆ EnableBreakOnFirstNaN

◆ EnableDetailedNanDetection

◆ EnableLayerDebugging

◆ EnableSingleStep

◆ EnableTesting

◆ forceSnapshot

◆ forceTest

◆ is_root_solver

◆ iter

◆ LabelQueryEpochs

◆ LabelQueryHitPercents

◆ LearningRateOverride

◆ MaximumIteration

◆ net

◆ parameter

◆ smoothed_loss

◆ SnapshotWeightUpdateMethod

◆ solver_count

◆ solver_rank

◆ Tag

◆ test_nets

◆ TestingIterationOverride

◆ TestingIterations

◆ TestingNet

◆ TrainingIterationOverride

◆ TrainingIterations