The Blob is the main holder of data that moves through the Layers of the Net. More...

Inheritance diagram for MyCaffe.common.Blob< T >:

Public Member Functions
	Blob (CudaDnn< T > cuda, Log log, bool bIncludeDiff=true, bool bUseHalfSize=false)
	The Blob constructor. More...

	Blob (CudaDnn< T > cuda, Log log, int nNum, int nChannels, int nHeight, int nWidth, bool bIncludeDiff=true, bool bUseHalfSize=false)
	DEPRECIATED; use More...

	Blob (CudaDnn< T > cuda, Log log, List< int > rgShape, bool bIncludeDiff=true, bool bUseHalfSize=false)
	The Blob constructor. More...

	Blob (CudaDnn< T > cuda, Log log, int[] rgShape, bool bIncludeDiff=true, bool bUseHalfSize=false)
	The Blob constructor. More...

	Blob (CudaDnn< T > cuda, Log log, Blob< T > b, bool bUseHalfSize=false)
	The Blob constructor. More...

	Blob (CudaDnn< T > cuda, Log log, SimpleDatum d, bool bCopyData=false, bool bIncludeDiff=true, bool bUseHalfSize=false)
	The Blob constructor. More...

	Blob (CudaDnn< T > cuda, Log log, BlobProto bp, bool bUseHalfSize=false)
	The Blob constructor. More...

	Blob (Blob< T > blob, long lCount, long lOffset)
	The Blob constructor used to copy another blob by creating memory pointers to its data thus sharing the same GPU memory. More...

void	Unsqueeze (int nNumAxes)
	Unsqueeze the shape by adding shape=1 on each axis until the 'nNumAxes' is reached. More...

double?	GetParameter (string strName)
	Get a blob parameter. More...

void	SetParameter (string strName, double dfVal)
	Set a blob parameter. More...

void	CopyParameters (Blob< T > b)
	Copy the parameters from another blob. More...

ulong	GetConversionWorkSize (bool bUseHalfSize)
	Returns the amount of memory (in bytes) required to convert from base to half and back. More...

void	ConvertToHalf (long hWorkMem, ulong lWorkSize, bool bData, bool bDiff)
	Converts this blob from its base type to the half type. More...

void	ConvertToBase (long hWorkMem, ulong lWorkSize, bool bData, bool bDiff)
	Converts this blob from the half type to the base type. More...

void	Dispose ()
	Releases all resources used by the Blob (including both GPU and Host). More...

void	Reshape (int nNum, int nChannels, int nHeight, int nWidth, bool? bUseHalfSize=null)
	DEPRECIATED; use More...

void	Reshape (List< int > rgShape, bool? bUseHalfSize=null)
	Change the dimensions of the blob, allocating new memory if necessary. More...

void	Reshape (int[] rgShape, bool? bUseHalfSize=null)
	Change the dimensions of the blob, allocating new memory if necessary. More...

void	Reshape (BlobShape shape, bool? bUseHalfSize=null)
	Change the dimensions of the blob, allocating new memory if necessary. More...

void	ReshapeLike (Blob< T > b, bool? bUseHalfSize=null)
	Reshape this Blob to have the same shape as another Blob. More...

List< int >	shape ()
	Returns an array where each element contains the shape of an axis of the Blob. More...

int	shape (int nIdx)
	Returns the dimension of the nIdx'th axis (or the negative nIdx'th axis from teh end, if nIdx is negative. More...

int	count ()
	Returns the total number of items in the Blob. More...

int	count (int nStartIdx, int nEndIdx)
	Compute the volume of a slice; i.e., the product of dimensions among a range of axes. More...

int	count (int nStartIdx)
	Compute the volume of a slice spanning from a particular first axis to the final axis. More...

int	CanonicalAxisIndex (int nIdx)
	Returns the 'canonical' version of a (usually) user-specified axis, allowing for negative indexing (e.g., -1 for the last axis). More...

int	LegacyShape (int nIdx)
	Returns the legacy shape at a given axis. More...

int	offset (int n, int c=0, int h=0, int w=0)
	Returns the flat offset given the number, channel, height and width. More...

int	offset (List< int > rgIdx)
	Returns the flat offset given the array of axes values. More...

void	CopyFrom (Blob< T > src, int nSrcOffset, int nDstOffset, int nCount, bool bCopyData, bool bCopyDiff)
	Copy from a source Blob. More...

long	CopyFrom (Blob< T > src, bool bCopyDiff=false, bool bReshape=false, long hDstHostBuffer=0, bool bIgnoreShape=false)
	Copy from a source Blob. More...

void	CopyFromAndPad (Blob< T > src, double dfPad=0, bool bCopyDiff=false)
	Copy the source data to this Blob, and if this blob is larger than the source, pad this blob with 'dfPad' values to the right. More...

void	CopyFromAndTransposeHeightWidth (Blob< T > blobSrc, bool bCopyDiff=false, bool bUseCuda=true)
	Copy from a source Blob and transpose the height and width of the copy. More...

void	CopyFrom (Blob< T > blobSrc, int nChannelFrom, int nChannelTo, bool bCopyDiff=false)
	Copy all data along a given channel from the source. More...

bool	Compare (Blob< T > other, Blob< T > work, bool bDiff=false, double dfTol=1e-8, bool bZeroCheck=true, bool bFullCompare=false, bool bDetectNans=true, bool bForceOtherData=false)
	Compare the data (or diff) of one blob to another and return true if all items fall within the specified tolerance or not. More...

bool	CompareEx (Blob< T > other, Blob< T > work, out double dfMin, out double dfMax, bool bDiff=false, double dfTol=1e-8, bool bZeroCheck=true, bool bFullCompare=false, bool bDetectNans=true, bool bForceOtherData=false)
	Compare the data (or diff) of one blob to another and return true if all items fall within the specified tolerance or not. More...

bool	ValidateData (Blob< T > work, bool bDiff=false)
	Validate the data or diff looking for NAN or Inf. More...

bool	Compare (CudaDnn< double > cuda, Blob< T > other, Blob< double > work, bool bDiff=false, double dfTol=1e-8)
	Compare the data (or diff) of one blob to another and return true if all items fall within the specified tolerance or not. More...

T	data_at (int n, int c, int h, int w)
	Returns the data at a given location in the Blob. More...

T	diff_at (int n, int c, int h, int w)
	Returns the diff at a given location in the Blob. More...

T	data_at (List< int > rgIdx)
	Returns the data at a given location in the Blob. More...

T	diff_at (List< int > rgIdx)
	Returns the diff at a given location in the Blob. More...

T[]	update_cpu_data ()
	Update the CPU data by transferring the GPU data over to the Host. More...

T[]	update_cpu_diff ()
	Update the CPU diff by transferring the GPU diff over to the Host. More...

void	Update ()
	The 'update' method is used for parameter blobs in a Net. More...

void	FromProto (BlobProto bp, bool bReshape=true)
	Create a new Blob from a given BlobProto. More...

BlobProto	ToProto (bool bWriteDiff=false)
	Writes the Blob to a new BlobProto. More...

T	asum_data ()
	Compute the sum of absolute values (L1 norm) of the data. More...

T	asum_diff ()
	Compute the sum of absolute values (L1 norm) of the diff. More...

T	sumsq_data ()
	Calcualte the sum of squares (L2 norm squared) of the data. More...

T	sumsq_diff ()
	Calcualte the sum of squares (L2 norm squared) of the diff. More...

void	scale_data (double df)
	Scale the data by a scaling factor. More...

void	scale_diff (double df)
	Scale the diff by a scaling factor. More...

void	scale_to_range (double dfMin, double dfMax)
	Scale the data in the blob to the range [dfMin,dfMax]. More...

void	scale_data (T fScaleFactor)
	Scale the blob data by a constant factor. More...

void	scale_diff (T fScaleFactor)
	Scale the blob diff by a constant factor. More...

void	ShareData (Blob< T > b)
	Set the data to point to the data of the other blob – useful in Layers which simply perform a copy in their forward pass. More...

void	ShareDiff (Blob< T > b)
	Set the diff to point to the diff of the other blob – useful in Layers which simply perform a copy in their forward pass. More...

void	Share (Blob< T > b)
	Share another Blob with this one, by setting the data and diff to the same data and diff of the other Blob. More...

T	GetData (int nIdx)
	Returns the data at a given flat index within the Blob. More...

T	GetDiff (int nIdx)
	Returns the diff at a given flat index within the Blob. More...

void	SetData (T[] rgData, int nCount=-1, bool bSetCount=true)
	Sets a number of items within the Blob's data. More...

void	SetData (T fVal, int nIdx=-1)
	Either sets all of the data items in the Blob to a given value, or alternatively only sets a single indexed item to a given value. More...

void	SetData (double dfVal, int nIdx=-1)
	Either sets all of the data items in the Blob to a given value, or alternatively only sets a single indexed item to a given value. More...

void	SetData (double dfVal, int nStartIdx, int nCount)
	Set a data range with a given value. More...

void	SetDiff (double dfVal, int nIdx=-1)
	Either sets all of the diff items in the Blob to a given value, or alternatively only sets a single indexed item to a given value. More...

void	SetDiff (double dfVal, int nStartIdx, int nCount)
	Set a diff range with a given value. More...

void	SetDiff (T[] rgDiff, int nCount=-1, bool bSetCount=true)
	Sets a number of items within the Blob's diff. More...

void	SetData (SimpleDatum d, bool bReshape, bool bCopyData=true)
	Sets the Blob values to the data contained within a SimpleDatum. More...

void	SetCPUData (T[] rg)
	Sets just the CPU data to the data specified. More...

void	AsyncGpuPush (long hStream)
	Asynchronously pushes the CPU data, previously set with SetCPUData, to the GPU. More...

bool	ShapeEquals (BlobProto bp)
	Compares the shape of this blob to the shape within a BlobProto. More...

bool	CompareShape (List< int > rgShape, bool bCompareCpuDataLen=false)
	Compares the shape of this blob to another shape. More...

bool	CompareShape (int[] rgShape, bool bCompareCpuDataLen=false)
	Compares the shape of this blob to another shape. More...

string	ToSizeString ()
	Returns a string describing the 4D shape of the Blob. More...

Datum	ToDatum ()
	Returns a new Datum that contains the shape and data of the Blob. More...

Blob< T >	Clone ()
	Copies the Blob, including its data and diff. More...

Blob< T >	MathAdd (Blob< T > blobA, T fScale)
	Clones the input Blob, scales the clone and then adds the data from this Blob to it. More...

Blob< T >	MathSub (Blob< T > blobA)
	Clones the input Blob and subtracts the data from this blob from it. More...

Blob< T >	MathDiv (T fScale)
	Clones the input Blob and divides a scalar from all of the clones data items. More...

void	SaveBinary (string strFile, bool bData, bool bDiff, bool bIncludeName=true)
	Saves the blob to a binary file. More...

void	Save (BinaryWriter bw, bool bData, bool bDiff, bool bIncludeName=true)
	Saves this Blob to a binary stream. More...

byte[]	ToByteArray ()
	Saves this Blob to a byte array. More...

override string	ToString ()
	Returns a string representation of the Blob. More...

string	ToString (int nMax, bool bDiff=false)
	Get the string representation containing up to the first 'nMax' items. More...

double	GetMinData (out long lPos)
	Returns the minimum data and the position where the minimum is located in the data. More...

double	GetMaxData (out long lPos)
	Returns the maximum data and the position where the maximum is located in the data. More...

double	GetMinDiff (out long lPos)
	Returns the minimum diff and the position where the minimum is located in the diff. More...

double	GetMaxDiff (out long lPos)
	Returns the maximum diff and the position where the maximum is located in the diff. More...

Tuple< double, double, double, double >	minmax_data (Blob< T > work, bool bDetectNans=false, bool bUseChunks=false)
	Returns the minimum and maximum values in the data of the Blob. More...

Tuple< double, double, double, double >	minmax_diff (Blob< T > work, bool bDetectNans=false, bool bUseChunks=false)
	Returns the minimum and maximum values in the diff of the Blob. More...

void	add_scalar (double dfVal)
	Adds a scalar value to the Blob. More...

Blob< T >	Resize (List< int > rgShape)
	The Resize method resizes the 3rd and 4th axes of the blob. More...

void	NormalizeData (double? dfMean=null, double? dfStd=null)
	Normalize the blob data by subtracting the mean and dividing by the standard deviation. More...

double	mean (float[] rgDf=null, bool bDiff=false)
	Calculate the mean of the blob data. More...

double	sum (float[] rgDf=null, bool bDiff=false)
	Calculate the sum of the blob data. More...

double	std (double? dfMean=null, float[] rgDf=null)
	Calculate the standard deviation of the blob data. More...

Tuple< T, T, T >	SetPixel (int nX, int nY, byte R, byte G, byte B, TransformationParameter.COLOR_ORDER order=TransformationParameter.COLOR_ORDER.RGB)
	Set the values of a 3 channel image embedded within the blob in the order RGB or BGR where the full hxw is stored for each color in order. More...

Tuple< T, T, T >	SetPixel (int nX, int nY, Tuple< T, T, T > pixel, bool bReturnOriginal=false, TransformationParameter.COLOR_ORDER order=TransformationParameter.COLOR_ORDER.RGB, int nOffset=0)
	Sets a pixel to the values within a three item tuple where the first item is assigned RED, second GREEN and third BLUE. More...

void	SaveToImage (string strFile, bool bNonZeroExistOnly=true, bool bSaveDiff=false, Dictionary< float, Color > rgSpecialValues=null)
	Save the Blob to an image where values less than 0 are colored red, and values greater than 0 are colored green. Values = 0 are colored black. More...

void	SaveToNumpy (string strFile, bool bSaveDiff=false)
	Save a blob with data to a Numpy .npy file. More...

Tuple< float[], int[]>	LoadFromNumpy (string strFile, bool bLoadDiff=false, bool bLoadDataOnly=false, Log log=null, int nMax=int.MaxValue)
	Load a blob with data from a Numpy array .npy file. More...

void	MatMul (Blob< T > blobA, Blob< T > blobB, bool bReshape=false, bool bTransA=false, bool bTransB=false, double dfScale=1.0, bool bADiff=false, bool bBDiff=false, bool bCDiff=false)
	MatMul blobA with blobB and place the result in this blob (e.g. this = matmul(A, B)). All blobs are in row-major format. More...

void	MatMulGrad (Blob< T > blobA, Blob< T > blobB, Blob< T > blobWork, double dfScale=1.0)
	Calculates and propagates the gradient for blobA and BlobB given the input gradient in this blob's diff values. After this call, the blobA and blobB diff values are filled with their respective gradients. More...

void	Percentile (Blob< T > blobY, double dfPercentile)
	Calculates the percentile and places the result in blobY. More...

Static Public Member Functions
static Blob< T >	LoadBinary (CudaDnn< T > cuda, Log log, string strFile, bool bData, bool bDiff)
	Loads a blob from a binary file (previously saved with SaveBinary). More...

static Blob< T >	Load (CudaDnn< T > cuda, Log log, BinaryReader br, bool bData, bool bDiff)
	Lods a new Blob from a binary reader. More...

static Blob< T >	FromByteArray (CudaDnn< T > cuda, Log log, byte[] rg)
	A new Blob is created from the byte array, previously saved with ToByteArray. More...

static void	SaveToNumpy (string strFile, float[] rgData, int[] rgShape)
	Save a blob with data to a Numpy .npy file. More...

static void	SaveToNumpy (string strFile, int[] rgData, int[] rgShape)
	Save a blob with data to a Numpy .npy file. More...

static void	SaveToNumpy (string strFile, long[] rgData, int[] rgShape)
	Save a blob with data to a Numpy .npy file. More...

static Tuple< List< float[]>, int[], List< string > >	LoadFromNumpy (string strFile, Log log=null, int nMax=int.MaxValue, int nStartIdx=0, int nCount=int.MaxValue)
	Load the long, int, bool, float, or double data from a very large Numpy array .npy file, reading in col order. More...

Static Public Attributes
const int	MAX_BLOB_AXES = 32
	Defines the maximum number of Axes supported by the Blob. More...

Protected Member Functions
virtual void	Dispose (bool bDisposing)
	Releases all resources used by the Blob (including both GPU and Host). More...

Properties
static T	Zero `[get]`
	Returns Zero (0) in type T. More...

static T	One `[get]`
	Returns One (1) in type T. More...

static T	MinusOne `[get]`
	Returns MinusOne (-1) in type T. More...

bool	Padded `[getset]`
	Get/set the padding state of the blob. More...

bool	HalfSize `[get]`
	Returns whether or not this blob is using half sizes. More...

bool	freeze_learning `[getset]`
	Specifies whether or not the diff is applied to the data during Update. When freeze learning = true, the update is skipped. More...

CudaDnn< T >	Cuda `[get]`
	Returns the CudaDnn object that manages the Blob's memory."/> More...

Log	Log `[get]`
	Returns the Log associated with the blob. More...

string	shape_string `[get]`
	Returns a string describing the Blob's shape. More...

bool	DiffExists `[get]`
	Returns whether or not the Diff portion exists. More...

int	num_axes `[get]`
	Returns the number of axes in the Blob. More...

int	num_true_axes `[get]`
	Returns the number of true axes, ignoring the trailing ones. More...

int	num `[get]`
	DEPRECIATED; legacy shape accessor num: use shape(0) instead. More...

int	channels `[get]`
	DEPRECIATED; legacy shape accessor channels: use shape(1) instead. More...

int	height `[get]`
	DEPRECIATED; legacy shape accessor height: use shape(2) instead. More...

int	width `[get]`
	DEPRECIATED; legacy shape accessor width: use shape(3) instead. More...

SyncedMemory< T >	data `[get]`
	Returns the SyncedMemory that stores the data. More...

SyncedMemory< T >	diff `[get]`
	Returns the SyncedMemory that stores the diff. More...

T[]	cpu_data `[get]`
	Returns the last host data retrieved from the GPU. More...

T[]	mutable_cpu_data `[getset]`
	Get data from the GPU and bring it over to the host, or Set data from the Host and send it over to the GPU. More...

long	gpu_data `[get]`
	Returns the data GPU handle used by the CudaDnn connection. More...

long	mutable_gpu_data `[get]`
	Returns the data GPU handle used by the CudaDnn connection. More...

T[]	cpu_diff `[get]`
	Returns the last host diff retrieved from the GPU. More...

T[]	mutable_cpu_diff `[getset]`
	Get diff from the GPU and bring it over to the host, or Set diff from the Host and send it over to the GPU. More...

long	gpu_diff `[get]`
	Returns the diff GPU handle used by the CudaDnn connection. More...

long	mutable_gpu_diff `[get]`
	Returns the diff GPU handle used by the CudaDnn connection. More...

long	gpu_shape `[get]`
	Returns the shape GPU handle used by the CudaDnn connection. The shape data contains the shape information of the Blob for use in GPU operations. More...

bool	reshape_when_sharing `[getset]`
	When true, this Blob is reshaped to the source when sharing the source data (default = false). More...

bool	snapshot_requested `[getset]`
	Get/set the snapshot request. More...

string	Name `[getset]`
	Get/set the name of the Blob. More...

double	min_data `[get]`
	Returns the minimum value in the data of the Blob. More...

double	max_data `[get]`
	Returns the maximum value in the data of the Blob. More...

double	min_diff `[get]`
	Returns the minimum value in the diff of the Blob. More...

double	max_diff `[get]`
	Returns the maximum value in the diff of the Blob. More...

BLOB_TYPE	type `[getset]`
	Returns the BLOB_TYPE of the Blob. More...

object	Tag `[getset]`
	Returns a user defined object associated with the Blob. More...

Detailed Description

The Blob is the main holder of data that moves through the Layers of the Net.

Each blob holds Data and optionally Diff where the data is passed through the Net Layers on each forward pass, and the Diff contains the errors passed backward through the Net Layers on the backward pass.

Template Parameters

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

Definition at line 24 of file Blob.cs.

Constructor & Destructor Documentation

◆ Blob() [1/8]

MyCaffe.common.Blob< T >.Blob	(	CudaDnn< T >	cuda,
		Log	log,
		bool	bIncludeDiff = `true`,
		bool	bUseHalfSize = `false`
	)

The Blob constructor.

Parameters

cuda	Specifies the CudaDnn instance used to communidate with Cuda.
log	Specifies the Log for output.
bIncludeDiff	Optionally, specifies whether or not to include (and allocate) the Diff data.
bUseHalfSize	Optionally, specifies to use half size (FP16) for both data and diff. This option is only available when using the float base type 'T'.

Definition at line 64 of file Blob.cs.

◆ Blob() [2/8]

MyCaffe.common.Blob< T >.Blob	(	CudaDnn< T >	cuda,
		Log	log,
		int	nNum,
		int	nChannels,
		int	nHeight,
		int	nWidth,
		bool	bIncludeDiff = `true`,
		bool	bUseHalfSize = `false`
	)

DEPRECIATED; use

Blob(...,rgShape) instead.

Parameters

cuda	Specifies the CudaDnn instance used to communidate with Cuda.
log	Specifies the Log for output.
nNum	Specifies the number of inputs.
nChannels	Specifies the number of channels per input.
nHeight	Specifies the height of each input.
nWidth	Specifies the width of each input.
bIncludeDiff	Optionally, specifies whether or not to include (and allocate) the Diff data.
bUseHalfSize	Optionally, specifies to use half size (FP16) for both data and diff. This option is only available when using the float base type 'T'.

Definition at line 98 of file Blob.cs.

◆ Blob() [3/8]

MyCaffe.common.Blob< T >.Blob	(	CudaDnn< T >	cuda,
		Log	log,
		List< int >	rgShape,
		bool	bIncludeDiff = `true`,
		bool	bUseHalfSize = `false`
	)

The Blob constructor.

Parameters

cuda	Specifies the CudaDnn instance used to communidate with Cuda.
log	Specifies the Log for output.
rgShape	Specifies the shape of each axis of the Blob.
bIncludeDiff	Optionally, specifies whether or not to include (and allocate) the Diff data.
bUseHalfSize	Optionally, specifies to use half size (FP16) for both data and diff. This option is only available when using the float base type 'T'.

Definition at line 115 of file Blob.cs.

◆ Blob() [4/8]

MyCaffe.common.Blob< T >.Blob	(	CudaDnn< T >	cuda,
		Log	log,
		int[]	rgShape,
		bool	bIncludeDiff = `true`,
		bool	bUseHalfSize = `false`
	)

The Blob constructor.

Parameters

cuda	Specifies the CudaDnn instance used to communidate with Cuda.
log	Specifies the Log for output.
rgShape	Specifies the shape of each axis of the Blob.
bIncludeDiff	Optionally, specifies whether or not to include (and allocate) the Diff data.
bUseHalfSize	Optionally, specifies to use half size (FP16) for both data and diff. This option is only available when using the float base type 'T'.

Definition at line 131 of file Blob.cs.

◆ Blob() [5/8]

MyCaffe.common.Blob< T >.Blob	(	CudaDnn< T >	cuda,
		Log	log,
		Blob< T >	b,
		bool	bUseHalfSize = `false`
	)

The Blob constructor.

Parameters

cuda	Specifies the CudaDnn instance used to communidate with Cuda.
log	Specifies the Log for output.
b	Create this blob to be like another Blob (e.g. same shape).
bUseHalfSize	Optionally, specifies to use half size (FP16) for both data and diff. This option is only available when using the float base type 'T'.

Definition at line 146 of file Blob.cs.

◆ Blob() [6/8]

MyCaffe.common.Blob< T >.Blob	(	CudaDnn< T >	cuda,
		Log	log,
		SimpleDatum	d,
		bool	bCopyData = `false`,
		bool	bIncludeDiff = `true`,
		bool	bUseHalfSize = `false`
	)

The Blob constructor.

Parameters

cuda	Specifies the CudaDnn instance used to communidate with Cuda.
log	Specifies the Log for output.
d	Specifies the datum for which the Blob is shaped to match.
bCopyData	Optionally, specifies whether or not to actually copy the data. When false, the shape is set, but no data is copied.
bIncludeDiff	Optionally, specifies whether or not to include (and allocate) the Diff data.
bUseHalfSize	Optionally, specifies to use half size (FP16) for both data and diff. This option is only available when using the float base type 'T'.

Definition at line 163 of file Blob.cs.

◆ Blob() [7/8]

MyCaffe.common.Blob< T >.Blob	(	CudaDnn< T >	cuda,
		Log	log,
		BlobProto	bp,
		bool	bUseHalfSize = `false`
	)

The Blob constructor.

Parameters

cuda	Specifies the CudaDnn instance used to communidate with Cuda.
log	Specifies the Log for output.
bp	Specifies the BlobProto used to load the Blob.
bUseHalfSize	Optionally, specifies to use half size (FP16) for both data and diff. This option is only available when using the float base type 'T'.

Definition at line 176 of file Blob.cs.

◆ Blob() [8/8]

MyCaffe.common.Blob< T >.Blob	(	Blob< T >	blob,
		long	lCount,
		long	lOffset
	)

The Blob constructor used to copy another blob by creating memory pointers to its data thus sharing the same GPU memory.

Parameters

blob	Specifies the blob whos data is to be shared.
lCount	Specifies the number of items to share.
lOffset	Specifies the offset into the blob where the shareing starts.

Definition at line 188 of file Blob.cs.

Member Function Documentation

◆ add_scalar()

void MyCaffe.common.Blob< T >.add_scalar ( double dfVal )

Adds a scalar value to the Blob.

Parameters

dfVal Specifies the scalar value.

Definition at line 2779 of file Blob.cs.

◆ asum_data()

T MyCaffe.common.Blob< T >.asum_data ( )

Compute the sum of absolute values (L1 norm) of the data.

Returns

Definition at line 1706 of file Blob.cs.

◆ asum_diff()

T MyCaffe.common.Blob< T >.asum_diff ( )

Compute the sum of absolute values (L1 norm) of the diff.

Returns

Definition at line 1718 of file Blob.cs.

◆ AsyncGpuPush()

void MyCaffe.common.Blob< T >.AsyncGpuPush ( long hStream )

Asynchronously pushes the CPU data, previously set with SetCPUData, to the GPU.

Parameters

hStream Specifies a handle to the Cuda Stream to use for synchronization.

Definition at line 2062 of file Blob.cs.

◆ CanonicalAxisIndex()

int MyCaffe.common.Blob< T >.CanonicalAxisIndex ( int nIdx )

Returns the 'canonical' version of a (usually) user-specified axis, allowing for negative indexing (e.g., -1 for the last axis).

Parameters

nIdx	The axis index.

Returns: The zero based index is returned.

Definition at line 780 of file Blob.cs.

◆ Clone()

Blob< T > MyCaffe.common.Blob< T >.Clone ( )

Copies the Blob, including its data and diff.

Returns: A new copy of the Blob is returned.

Definition at line 2202 of file Blob.cs.

◆ Compare() [1/2]

bool MyCaffe.common.Blob< T >.Compare	(	Blob< T >	other,
		Blob< T >	work,
		bool	bDiff = `false`,
		double	dfTol = `1e-8`,
		bool	bZeroCheck = `true`,
		bool	bFullCompare = `false`,
		bool	bDetectNans = `true`,
		bool	bForceOtherData = `false`
	)

Compare the data (or diff) of one blob to another and return true if all items fall within the specified tolerance or not.

Parameters

other	Specifies the other blob to compare.
work	Specifies a temporary work blob.
bDiff	Specifies to compare the diff.
dfTol	Specifies the accepted tolerance.
bZeroCheck	Optionally, specifies to check for all zeros (default = false).
bFullCompare	Optionally, compare each item individually (default = false).
bDetectNans	Optionally, detect NAN's (default = true).
bForceOtherData	Optionally, force using the other blobs data even when bDiff = true.

Returns: If all data (or diff) values fall within the tolerance, true is returned, otherwise false.

Definition at line 1093 of file Blob.cs.

◆ Compare() [2/2]

bool MyCaffe.common.Blob< T >.Compare	(	CudaDnn< double >	cuda,
		Blob< T >	other,
		Blob< double >	work,
		bool	bDiff = `false`,
		double	dfTol = `1e-8`
	)

Compare the data (or diff) of one blob to another and return true if all items fall within the specified tolerance or not.

Parameters

cuda	Specifies a double precision CudaDnn.
other	Specifies the other blob to compare.
work	Specifies a temporary work blob.
bDiff	Specifies to compare the diff.
dfTol	Specifies the accepted tolerance.

Returns: If all data (or diff) values fall within the tolerance, true is returned, otherwise false.

Definition at line 1220 of file Blob.cs.

◆ CompareEx()

bool MyCaffe.common.Blob< T >.CompareEx	(	Blob< T >	other,
		Blob< T >	work,
		out double	dfMin,
		out double	dfMax,
		bool	bDiff = `false`,
		double	dfTol = `1e-8`,
		bool	bZeroCheck = `true`,
		bool	bFullCompare = `false`,
		bool	bDetectNans = `true`,
		bool	bForceOtherData = `false`
	)

Compare the data (or diff) of one blob to another and return true if all items fall within the specified tolerance or not.

Parameters

other	Specifies the other blob to compare.
work	Specifies a temporary work blob.
dfMin	Returns the min difference.
dfMax	Returns the max difference.
bDiff	Specifies to compare the diff.
dfTol	Specifies the accepted tolerance.
bZeroCheck	Optionally, specifies to check for all zeros (default = false).
bFullCompare	Optionally, compare each item individually (default = false).
bDetectNans	Optionally, detect NAN's (default = true).
bForceOtherData	Optionally, force using the other blobs data even when bDiff = true.

Returns: If all data (or diff) values fall within the tolerance, true is returned, otherwise false.

Definition at line 1114 of file Blob.cs.

◆ CompareShape() [1/2]

bool MyCaffe.common.Blob< T >.CompareShape	(	int[]	rgShape,
		bool	bCompareCpuDataLen = `false`
	)

Compares the shape of this blob to another shape.

Parameters

rgShape	Specifies the shape to compare with.
bCompareCpuDataLen	Optionally, compare the CPU data length.

Returns: If the shapes are the same, return true, othewise false.

Definition at line 2148 of file Blob.cs.

◆ CompareShape() [2/2]

bool MyCaffe.common.Blob< T >.CompareShape	(	List< int >	rgShape,
		bool	bCompareCpuDataLen = `false`
	)

Compares the shape of this blob to another shape.

Parameters

rgShape	Specifies the shape to compare with.
bCompareCpuDataLen	Optionally, compare the CPU data length.

Returns: If the shapes are the same, return true, othewise false.

Definition at line 2108 of file Blob.cs.

◆ ConvertToBase()

void MyCaffe.common.Blob< T >.ConvertToBase	(	long	hWorkMem,
		ulong	lWorkSize,
		bool	bData,
		bool	bDiff
	)

Converts this blob from the half type to the base type.

Parameters

hWorkMem	Specifies the work memory.
lWorkSize	Specifies the work size.
bData	Specifies to convert the data.
bDiff	Specifies to convert the diff

Definition at line 339 of file Blob.cs.

◆ ConvertToHalf()

void MyCaffe.common.Blob< T >.ConvertToHalf	(	long	hWorkMem,
		ulong	lWorkSize,
		bool	bData,
		bool	bDiff
	)

Converts this blob from its base type to the half type.

Parameters

hWorkMem	Specifies the work memory.
lWorkSize	Specifies the work size.
bData	Specifies to convert the data.
bDiff	Specifies to convert the diff

Definition at line 306 of file Blob.cs.

◆ CopyFrom() [1/3]

void MyCaffe.common.Blob< T >.CopyFrom	(	Blob< T >	blobSrc,
		int	nChannelFrom,
		int	nChannelTo,
		bool	bCopyDiff = `false`
	)

Copy all data along a given channel from the source.

Parameters

blobSrc	Specifies the source blob to copy from.
nChannelFrom	Specifies the channel within the source to copy from.
nChannelTo	Specifies the channel in the destination (this blob) to copy into.
bCopyDiff	Specifies whether or not to copy the data or diff.

Definition at line 1054 of file Blob.cs.

◆ CopyFrom() [2/3]

long MyCaffe.common.Blob< T >.CopyFrom	(	Blob< T >	src,
		bool	bCopyDiff = `false`,
		bool	bReshape = `false`,
		long	hDstHostBuffer = `0`,
		bool	bIgnoreShape = `false`
	)

Copy from a source Blob.

Parameters

src	The Blob to copy from.
bCopyDiff	If false, copy the data; if true, copy the diff.
bReshape	If false, require this Blob to be pre-shaped to the shape of other (and die otherwise); If true, Reshape this Blob to other's shape if necessary.
hDstHostBuffer	Optionally, specifies the host buffer of the destination.
bIgnoreShape	Optionally, specifies to ignore the shape and just make sure the count is the same before copying (default = false).

Returns: When used, the host buffer handle is returned.

Definition at line 928 of file Blob.cs.

◆ CopyFrom() [3/3]

void MyCaffe.common.Blob< T >.CopyFrom	(	Blob< T >	src,
		int	nSrcOffset,
		int	nDstOffset,
		int	nCount,
		bool	bCopyData,
		bool	bCopyDiff
	)

Copy from a source Blob.

Parameters

src	The Blob to copy from.
nSrcOffset	The offset into the source data to copy.
nDstOffset	The offset into the destination data to copy.
nCount	The number of items to copy.
bCopyData	Copy the data.
bCopyDiff	Copy the diff.

Definition at line 903 of file Blob.cs.

◆ CopyFromAndPad()

void MyCaffe.common.Blob< T >.CopyFromAndPad	(	Blob< T >	src,
		double	dfPad = `0`,
		bool	bCopyDiff = `false`
	)

Copy the source data to this Blob, and if this blob is larger than the source, pad this blob with 'dfPad' values to the right.

Parameters

src	Specifies the source Blob to copy.
dfPad	Specifies the pad value (default = 0).
bCopyDiff	Specifies to copy the diff values (default = false).

Definition at line 973 of file Blob.cs.

◆ CopyFromAndTransposeHeightWidth()

void MyCaffe.common.Blob< T >.CopyFromAndTransposeHeightWidth	(	Blob< T >	blobSrc,
		bool	bCopyDiff = `false`,
		bool	bUseCuda = `true`
	)

Copy from a source Blob and transpose the height and width of the copy.

Parameters

blobSrc	Specifies the Blob to copy from.
bCopyDiff	Optionally, specifies to copy and transform the diff instead of the data (default = false).
bUseCuda	Optionally, specifies to use CUDA function for transformations (default = true)

Definition at line 1002 of file Blob.cs.

◆ CopyParameters()

void MyCaffe.common.Blob< T >.CopyParameters ( Blob< T > b )

Copy the parameters from another blob.

Parameters

b	Specifies the blob whose parameters are to be copied.

Definition at line 245 of file Blob.cs.

◆ count() [1/3]

int MyCaffe.common.Blob< T >.count ( )

Returns the total number of items in the Blob.

Returns

Definition at line 739 of file Blob.cs.

◆ count() [2/3]

int MyCaffe.common.Blob< T >.count ( int nStartIdx )

Compute the volume of a slice spanning from a particular first axis to the final axis.

Parameters

nStartIdx The first axis to include in the slice.

Returns

Definition at line 768 of file Blob.cs.

◆ count() [3/3]

int MyCaffe.common.Blob< T >.count	(	int	nStartIdx,
		int	nEndIdx
	)

Compute the volume of a slice; i.e., the product of dimensions among a range of axes.

Parameters

nStartIdx	The first axis to include in the slice.
nEndIdx	The first axis to exclude from the slice.

Returns

Definition at line 751 of file Blob.cs.

◆ data_at() [1/2]

T MyCaffe.common.Blob< T >.data_at	(	int	n,
		int	c,
		int	h,
		int	w
	)

Returns the data at a given location in the Blob.

Parameters

n	Specifies the num.
c	Specifies the channel.
h	Specifies the height.
w	Specifies the width.

Returns: The data item at the location is returned.

Definition at line 1394 of file Blob.cs.

◆ data_at() [2/2]

T MyCaffe.common.Blob< T >.data_at ( List< int > rgIdx )

Returns the data at a given location in the Blob.

Parameters

rgIdx Specifies an array of axes.

Returns: The data item at the location is returned.

Definition at line 1417 of file Blob.cs.

◆ diff_at() [1/2]

T MyCaffe.common.Blob< T >.diff_at	(	int	n,
		int	c,
		int	h,
		int	w
	)

Returns the diff at a given location in the Blob.

Parameters

n	Specifies the num.
c	Specifies the channel.
h	Specifies the height.
w	Specifies the width.

Returns: The diff item at the location is returned.

Definition at line 1407 of file Blob.cs.

◆ diff_at() [2/2]

T MyCaffe.common.Blob< T >.diff_at ( List< int > rgIdx )

Returns the diff at a given location in the Blob.

Parameters

rgIdx Specifies an array of axes.

Returns: The diff item at the location is returned.

Definition at line 1427 of file Blob.cs.

◆ Dispose() [1/2]

void MyCaffe.common.Blob< T >.Dispose ( )

Releases all resources used by the Blob (including both GPU and Host).

Definition at line 429 of file Blob.cs.

◆ Dispose() [2/2]

virtual void MyCaffe.common.Blob< T >.Dispose ( bool bDisposing )

protectedvirtual

Releases all resources used by the Blob (including both GPU and Host).

Parameters

bDisposing Set to true when disposing the object.

Definition at line 402 of file Blob.cs.

◆ FromByteArray()

static Blob< T > MyCaffe.common.Blob< T >.FromByteArray	(	CudaDnn< T >	cuda,
		Log	log,
		byte[]	rg
	)

static

A new Blob is created from the byte array, previously saved with ToByteArray.

Parameters

cuda	Specifies the cuda connection.
log	Specifies the output log.
rg	Specifies the byte array to convert.

Returns: The new Blob loaded with the data in the byte array is returned.

Definition at line 2453 of file Blob.cs.

◆ FromProto()

void MyCaffe.common.Blob< T >.FromProto	(	BlobProto	bp,
		bool	bReshape = `true`
	)

Create a new Blob from a given BlobProto.

Parameters

bp	Specifies the BlobProto to use.
bReshape	Specifies whether or not to reshape the Blob with the BlobProto shape.

Definition at line 1589 of file Blob.cs.

◆ GetConversionWorkSize()

ulong MyCaffe.common.Blob< T >.GetConversionWorkSize ( bool bUseHalfSize )

Returns the amount of memory (in bytes) required to convert from base to half and back.

Parameters

bUseHalfSize Specifies whether or not we are converting to half size or not.

Definition at line 293 of file Blob.cs.

◆ GetData()

T MyCaffe.common.Blob< T >.GetData ( int nIdx )

Returns the data at a given flat index within the Blob.

Parameters

nIdx	Specifies the flat index in the range [0,count()-1].

Returns: The data item at the index is returned.

Definition at line 1893 of file Blob.cs.

◆ GetDiff()

T MyCaffe.common.Blob< T >.GetDiff ( int nIdx )

Returns the diff at a given flat index within the Blob.

Parameters

nIdx	Specifies the flat index in the range [0,count()-1].

Returns: The diff item at the index is returned.

Definition at line 1907 of file Blob.cs.

◆ GetMaxData()

double MyCaffe.common.Blob< T >.GetMaxData ( out long lPos )

Returns the maximum data and the position where the maximum is located in the data.

Parameters

lPos	Returns the location of the maximum.

Returns: Returns the maximum value.

Definition at line 2538 of file Blob.cs.

◆ GetMaxDiff()

double MyCaffe.common.Blob< T >.GetMaxDiff ( out long lPos )

Returns the maximum diff and the position where the maximum is located in the diff.

Parameters

lPos	Returns the location of the maximum.

Returns: Returns the maximum value.

Definition at line 2590 of file Blob.cs.

◆ GetMinData()

double MyCaffe.common.Blob< T >.GetMinData ( out long lPos )

Returns the minimum data and the position where the minimum is located in the data.

Parameters

lPos	Returns the location of the minimum.

Returns: Returns the minimum value.

Definition at line 2512 of file Blob.cs.

◆ GetMinDiff()

double MyCaffe.common.Blob< T >.GetMinDiff ( out long lPos )

Returns the minimum diff and the position where the minimum is located in the diff.

Parameters

lPos	Returns the location of the minimum.

Returns: Returns the minimum value.

Definition at line 2564 of file Blob.cs.

◆ GetParameter()

double? MyCaffe.common.Blob< T >.GetParameter ( string strName )

Get a blob parameter.

Parameters

strName Specifies the name of the blob parameter.

Returns: If the parameter exists it is returned, otherwise null is returned.

Definition at line 220 of file Blob.cs.

◆ LegacyShape()

int MyCaffe.common.Blob< T >.LegacyShape ( int nIdx )

Returns the legacy shape at a given axis.

Parameters

nIdx	Specifies the axis.

Returns: The shape at a given axis is returned.

Definition at line 825 of file Blob.cs.

◆ Load()

static Blob< T > MyCaffe.common.Blob< T >.Load	(	CudaDnn< T >	cuda,
		Log	log,
		BinaryReader	br,
		bool	bData,
		bool	bDiff
	)

static

Lods a new Blob from a binary reader.

Parameters

cuda	Specifies the instance of the CudaDnn connection to use.
log	Specifies the Log for output.
br	Specifies the binary reader.
bData	Specifies whether or not to read in the data.
bDiff	Specifies whether or not to read in the diff.

Returns

Definition at line 2377 of file Blob.cs.

◆ LoadBinary()

static Blob< T > MyCaffe.common.Blob< T >.LoadBinary	(	CudaDnn< T >	cuda,
		Log	log,
		string	strFile,
		bool	bData,
		bool	bDiff
	)

static

Loads a blob from a binary file (previously saved with SaveBinary).

Parameters

cuda	Specifies the instance of the CudaDnn connection to use.
log	Specifies the Log for output.
strFile	Specifies the file to load the data/diff from.
bData	Specifies to load the data.
bDiff	Specifies to load the diff.

Definition at line 2359 of file Blob.cs.

◆ LoadFromNumpy() [1/2]

Tuple< float[], int[]> MyCaffe.common.Blob< T >.LoadFromNumpy	(	string	strFile,
		bool	bLoadDiff = `false`,
		bool	bLoadDataOnly = `false`,
		Log	log = `null`,
		int	nMax = `int.MaxValue`
	)

Load a blob with data from a Numpy array .npy file.

Parameters

strFile	Specifies the .npy file name.
bLoadDiff	Specifies to load the diff, when false, the data is loaded.
bLoadDataOnly	Specifies to load the data and return it as an array but do not load the gpu memory.
log	Optionally, specifies the output log.
nMax	Optionally, specifies the maximum number of items to load.

Exceptions

Exception An exception is thrown when an invalid or unsupported feature is located.

See also: A Simple File Format for NumPy Arrays

Returns: A tuple containing the float[] data and int[] shape is returned.

Definition at line 3473 of file Blob.cs.

◆ LoadFromNumpy() [2/2]

static Tuple< List< float[]>, int[], List< string > > MyCaffe.common.Blob< T >.LoadFromNumpy	(	string	strFile,
		Log	log = `null`,
		int	nMax = `int.MaxValue`,
		int	nStartIdx = `0`,
		int	nCount = `int.MaxValue`
	)

static

Load the long, int, bool, float, or double data from a very large Numpy array .npy file, reading in col order.

Parameters

strFile	Specifies the .npy file name.
log	Optionally, specifies the output log.
nMax	Optionally, specifies the maximum number of items to load (default = MaxValue).
nStartIdx	Optionally, specifies a start index of items to read (default = 0).
nCount	Optionally, specifies the number of items to load from the start index (default = MaxValue).

Exceptions

Exception An exception is thrown when an invalid or unsupported feature is located.

See also: A Simple File Format for NumPy Arrays Note, when loading at a given start index, only the 'nCount' number of items are loaded from the start or the file is read to the end, whichever is sooner.

Returns: A tuple containing the float[,] data and int[] shape is returned.

Definition at line 3578 of file Blob.cs.

◆ MathAdd()

Blob< T > MyCaffe.common.Blob< T >.MathAdd	(	Blob< T >	blobA,
		T	fScale
	)

Clones the input Blob, scales the clone and then adds the data from this Blob to it.

Calculation: Y = Clone(blobA) * dfScale + this

Parameters

blobA	Specifies the Blob to clone.
fScale	Specifies the scaling factor to apply to the clone.

Returns: The Blob copy is returned.

Definition at line 2227 of file Blob.cs.

◆ MathDiv()

Blob< T > MyCaffe.common.Blob< T >.MathDiv ( T fScale )

Clones the input Blob and divides a scalar from all of the clones data items.

Calculation: Y = Clone(this) * fScale

Parameters

fScale Specifies scaling factor.

Returns: The Blob copy is returned.

Definition at line 2264 of file Blob.cs.

◆ MathSub()

Blob< T > MyCaffe.common.Blob< T >.MathSub ( Blob< T > blobA )

Clones the input Blob and subtracts the data from this blob from it.

Calculation: Y = Clone(blobA) + this

Parameters

blobA Specifies the Blob to clone.

Returns: The Blob copy is returned.

Definition at line 2247 of file Blob.cs.

◆ MatMul()

void MyCaffe.common.Blob< T >.MatMul	(	Blob< T >	blobA,
		Blob< T >	blobB,
		bool	bReshape = `false`,
		bool	bTransA = `false`,
		bool	bTransB = `false`,
		double	dfScale = `1.0`,
		bool	bADiff = `false`,
		bool	bBDiff = `false`,
		bool	bCDiff = `false`
	)

MatMul blobA with blobB and place the result in this blob (e.g. this = matmul(A, B)). All blobs are in row-major format.

Parameters

blobA	Specifies the first input with last 2 axes size of MxK row-major matrix (first axes must match blobB's)
blobB	Specifies the second input with last 2 axes size of KxN row-major matrix (first axes must match blobA's)
dfScale	Specifies the scale applied to blobB.
bReshape	Specifies to reshape this blob the the expected shape (default = false).
bTransA	Specifies to transpose A first.
bTransB	Specifies to transpose B first.
bADiff	Specifies to use the diff values in blobA, otherwise the data values are used (default = false).
bBDiff	Specifies to use the diff values in blobB, otherwise the data values are used (default = false).
bCDiff	Specifies to use the diff values in blobC, otherwise the data values are used (default = false).

Definition at line 3922 of file Blob.cs.

◆ MatMulGrad()

void MyCaffe.common.Blob< T >.MatMulGrad	(	Blob< T >	blobA,
		Blob< T >	blobB,
		Blob< T >	blobWork,
		double	dfScale = `1.0`
	)

Calculates and propagates the gradient for blobA and BlobB given the input gradient in this blob's diff values. After this call, the blobA and blobB diff values are filled with their respective gradients.

Parameters

blobA	Specifies the blobA where blobA gradients are placed (in diff values).
blobB	Specifies the blobB where blobB gradients are placed (in diff values).
blobWork	Specifies a work blob.
dfScale	Specifies a scale to be applied to the diffs in this blob before the MatMul (default = 1.0).

See also: PyGrad:functions.py by Jake Tae, 2020, GitHub:jaketae/pygrad

Definition at line 3974 of file Blob.cs.

◆ mean()

double MyCaffe.common.Blob< T >.mean	(	float[]	rgDf = `null`,
		bool	bDiff = `false`
	)

Calculate the mean of the blob data.

Parameters

rgDf	Optionally, specifies the CPU data to use (default = null).
bDiff	Optionally, specifies to use the diff instead of the data.

Returns: The mean is returned.

Definition at line 2965 of file Blob.cs.

◆ minmax_data()

Tuple< double, double, double, double > MyCaffe.common.Blob< T >.minmax_data	(	Blob< T >	work,
		bool	bDetectNans = `false`,
		bool	bUseChunks = `false`
	)

Returns the minimum and maximum values in the data of the Blob.

Parameters

work	Specifies a workspace used to optimize the query.
bDetectNans	Optionally, specifies whether or not to detect Nan's and Infinity values.
bUseChunks	Optionally, specifies whether or not to use the min/max on all of the memory (default) or run in chunks (true).

Returns: A tuple containing the 'min', 'max' and optionally 'number of nans' and 'number of infinity' is returned for the data.

Definition at line 2624 of file Blob.cs.

◆ minmax_diff()

Tuple< double, double, double, double > MyCaffe.common.Blob< T >.minmax_diff	(	Blob< T >	work,
		bool	bDetectNans = `false`,
		bool	bUseChunks = `false`
	)

Returns the minimum and maximum values in the diff of the Blob.

Parameters

work	Specifies a workspace used to optimize the query.
bDetectNans	Optionally, specifies whether or not to detect Nan's and Infinity values.
bUseChunks	Optionally, specifies whether or not to use the min/max on all of the memory (default) or run in chunks (true).

Returns: A tuple containing the 'min', 'max' and optionally 'number of nans' and 'number of infinity' is returned for the data.

Definition at line 2694 of file Blob.cs.

◆ NormalizeData()

void MyCaffe.common.Blob< T >.NormalizeData	(	double?	dfMean = `null`,
		double?	dfStd = `null`
	)

Normalize the blob data by subtracting the mean and dividing by the standard deviation.

Parameters

dfMean	Optionally, specifies a mean value to use (default = null).
dfStd	Optionally, specifies a std value to use (default = null).

Definition at line 2942 of file Blob.cs.

◆ offset() [1/2]

int MyCaffe.common.Blob< T >.offset	(	int	n,
		int	c = `0`,
		int	h = `0`,
		int	w = `0`
	)

Returns the flat offset given the number, channel, height and width.

Parameters

n	Specifies the num.
c	Specifies the channel.
h	Specifies the height.
w	Specifies the width.

Returns: The flat offset is returned.

Definition at line 850 of file Blob.cs.

◆ offset() [2/2]

int MyCaffe.common.Blob< T >.offset ( List< int > rgIdx )

Returns the flat offset given the array of axes values.

Parameters

rgIdx Specifies the array of axes values.

Returns: The flat offset is returned.

Definition at line 873 of file Blob.cs.

◆ Percentile()

void MyCaffe.common.Blob< T >.Percentile	(	Blob< T >	blobY,
		double	dfPercentile
	)

Calculates the percentile and places the result in blobY.

BlobY is reshaped to size 1xCxH*W.

Parameters

blobY	Specifies blob where the percentile results are placed.
dfPercentile	Specifies the percentile specified in range (0,1].

Definition at line 3992 of file Blob.cs.

◆ Reshape() [1/4]

void MyCaffe.common.Blob< T >.Reshape	(	BlobShape	shape,
		bool?	bUseHalfSize = `null`
	)

Change the dimensions of the blob, allocating new memory if necessary.

This function can be called both to create an initial allocation of memory, and to adjust the dimensions of a top blob during Layer::Reshape or Layer::Forward. When changing the size of blob, memory will only be reallocated if sufficient memory does not already exist, and excess memory will not be freed until Dispose is called.

Note that reshaping an input blob and immediately calling Net::Backward is an error; either Net::Forward or Net::Reshape need to be called to propagate the new input shape to higher layers.

Parameters

shape	Specifies the new shape of the Blob.
bUseHalfSize	Optionally, specifies to use half sized memory.

Definition at line 637 of file Blob.cs.

◆ Reshape() [2/4]

void MyCaffe.common.Blob< T >.Reshape	(	int	nNum,
		int	nChannels,
		int	nHeight,
		int	nWidth,
		bool?	bUseHalfSize = `null`
	)

DEPRECIATED; use

Reshape(rgShape) instead.

Parameters

nNum
nChannels
nHeight
nWidth
bUseHalfSize	Optionally, specifies to use half sized memory.

Definition at line 442 of file Blob.cs.

◆ Reshape() [3/4]

void MyCaffe.common.Blob< T >.Reshape	(	int[]	rgShape,
		bool?	bUseHalfSize = `null`
	)

Change the dimensions of the blob, allocating new memory if necessary.

This function can be called both to create an initial allocation of memory, and to adjust the dimensions of a top blob during Layer::Reshape or Layer::Forward. When changing the size of blob, memory will only be reallocated if sufficient memory does not already exist, and excess memory will not be freed until Dispose is called.

Note that reshaping an input blob and immediately calling Net::Backward is an error; either Net::Forward or Net::Reshape need to be called to propagate the new input shape to higher layers.

Parameters

rgShape	Specifies the new shape.
bUseHalfSize	Optionally, specifies to use half sized memory.

Definition at line 583 of file Blob.cs.

◆ Reshape() [4/4]

void MyCaffe.common.Blob< T >.Reshape	(	List< int >	rgShape,
		bool?	bUseHalfSize = `null`
	)

Change the dimensions of the blob, allocating new memory if necessary.

This function can be called both to create an initial allocation of memory, and to adjust the dimensions of a top blob during Layer::Reshape or Layer::Forward. When changing the size of blob, memory will only be reallocated if sufficient memory does not already exist, and excess memory will not be freed until Dispose is called.

Note that reshaping an input blob and immediately calling Net::Backward is an error; either Net::Forward or Net::Reshape need to be called to propagate the new input shape to higher layers.

Parameters

rgShape	Specifies the new shape.
bUseHalfSize	Optionally, specifies to use half sized memory.

Definition at line 562 of file Blob.cs.

◆ ReshapeLike()

void MyCaffe.common.Blob< T >.ReshapeLike	(	Blob< T >	b,
		bool?	bUseHalfSize = `null`
	)

Reshape this Blob to have the same shape as another Blob.

Parameters

b	Specifies the other Blob.
bUseHalfSize	Optionally, specifies to use half sized memory.

Definition at line 648 of file Blob.cs.

◆ Resize()

Blob< T > MyCaffe.common.Blob< T >.Resize ( List< int > rgShape )

The Resize method resizes the 3rd and 4th axes of the blob.

Currently, the Resize method only works on 4 axis blobs. Resize is different from Reshape in that Resize actually averages the data when resizing the blob.

Parameters

rgShape Specifies the new shape to resize to.

Returns: A newly re-sized Blob is returned.

Definition at line 2852 of file Blob.cs.

◆ Save()

void MyCaffe.common.Blob< T >.Save	(	BinaryWriter	bw,
		bool	bData,
		bool	bDiff,
		bool	bIncludeName = `true`
	)

Saves this Blob to a binary stream.

Parameters

bw	Specifies the binary writer.
bData	Specifies whether or not to save the data.
bDiff	Specifies whether or not to save the diff.
bIncludeName	Specifies whether or not to include the name.

Definition at line 2298 of file Blob.cs.

◆ SaveBinary()

void MyCaffe.common.Blob< T >.SaveBinary	(	string	strFile,
		bool	bData,
		bool	bDiff,
		bool	bIncludeName = `true`
	)

Saves the blob to a binary file.

Parameters

strFile	Specifies the file to save the data/diff into.
bData	Specifies to save the data.
bDiff	Specifies to save the diff.
bIncludeName	Specifies whether or not to include the name.

Definition at line 2282 of file Blob.cs.

◆ SaveToImage()

void MyCaffe.common.Blob< T >.SaveToImage	(	string	strFile,
		bool	bNonZeroExistOnly = `true`,
		bool	bSaveDiff = `false`,
		Dictionary< float, Color >	rgSpecialValues = `null`
	)

Save the Blob to an image where values less than 0 are colored red, and values greater than 0 are colored green. Values = 0 are colored black.

Parameters

strFile	Specifies the file where the image is saved.
bNonZeroExistOnly	Optionally, specifies to color whether data exists as nonZero only (default = true).
bSaveDiff	Optionally, specifies to save the diff instead of the data (default = false).
rgSpecialValues	Optionally, specifies special values with associated colors.

Definition at line 3105 of file Blob.cs.

◆ SaveToNumpy() [1/4]

void MyCaffe.common.Blob< T >.SaveToNumpy	(	string	strFile,
		bool	bSaveDiff = `false`
	)

Save a blob with data to a Numpy .npy file.

Parameters

strFile	Specifies the .npy file name where the data is saved.
bSaveDiff	Specifies to save the diff, when false, the data is saved.

See also: A Simple File Format for NumPy Arrays

Definition at line 3259 of file Blob.cs.

◆ SaveToNumpy() [2/4]

static void MyCaffe.common.Blob< T >.SaveToNumpy	(	string	strFile,
		float[]	rgData,
		int[]	rgShape
	)

static

Save a blob with data to a Numpy .npy file.

Parameters

strFile	Specifies the .npy file name where the data is saved.
rgData	Specifies the data to be saved.
rgShape	Specifies the shape of the data to be saved.

See also: A Simple File Format for NumPy Arrays

Definition at line 3316 of file Blob.cs.

◆ SaveToNumpy() [3/4]

static void MyCaffe.common.Blob< T >.SaveToNumpy	(	string	strFile,
		int[]	rgData,
		int[]	rgShape
	)

static

Save a blob with data to a Numpy .npy file.

Parameters

strFile	Specifies the .npy file name where the data is saved.
rgData	Specifies the data to be saved.
rgShape	Specifies the shape of the data to be saved.

See also: A Simple File Format for NumPy Arrays

Definition at line 3367 of file Blob.cs.

◆ SaveToNumpy() [4/4]

static void MyCaffe.common.Blob< T >.SaveToNumpy	(	string	strFile,
		long[]	rgData,
		int[]	rgShape
	)

static

Save a blob with data to a Numpy .npy file.

Parameters

strFile	Specifies the .npy file name where the data is saved.
rgData	Specifies the data to be saved.
rgShape	Specifies the shape of the data to be saved.

See also: A Simple File Format for NumPy Arrays

Definition at line 3418 of file Blob.cs.

◆ scale_data() [1/2]

void MyCaffe.common.Blob< T >.scale_data ( double df )

Scale the data by a scaling factor.

Parameters

df	Specifies the scaling factor.

Definition at line 1754 of file Blob.cs.

◆ scale_data() [2/2]

void MyCaffe.common.Blob< T >.scale_data ( T fScaleFactor )

Scale the blob data by a constant factor.

Parameters

fScaleFactor Specifies the scaling factor.

Definition at line 1782 of file Blob.cs.

◆ scale_diff() [1/2]

void MyCaffe.common.Blob< T >.scale_diff ( double df )

Scale the diff by a scaling factor.

Parameters

df	Specifies the scaling factor.

Definition at line 1763 of file Blob.cs.

◆ scale_diff() [2/2]

void MyCaffe.common.Blob< T >.scale_diff ( T fScaleFactor )

Scale the blob diff by a constant factor.

Parameters

fScaleFactor Specifies the scaling factor.

Definition at line 1791 of file Blob.cs.

◆ scale_to_range()

void MyCaffe.common.Blob< T >.scale_to_range	(	double	dfMin,
		double	dfMax
	)

Scale the data in the blob to the range [dfMin,dfMax].

Parameters

dfMin	Specifies the minimum of the range.
dfMax	Specifies the maximum of the range.

Definition at line 1773 of file Blob.cs.

◆ SetCPUData()

void MyCaffe.common.Blob< T >.SetCPUData ( T[] rg )

Sets just the CPU data to the data specified.

Parameters

rg	Specifies the CPU data to set.

Definition at line 2052 of file Blob.cs.

◆ SetData() [1/5]

void MyCaffe.common.Blob< T >.SetData	(	double	dfVal,
		int	nIdx = `-1`
	)

Either sets all of the data items in the Blob to a given value, or alternatively only sets a single indexed item to a given value.

Parameters

dfVal	Specifies the value to set.
nIdx	Optionally, specifies the index of the item to set.

Definition at line 1947 of file Blob.cs.

◆ SetData() [2/5]

void MyCaffe.common.Blob< T >.SetData	(	double	dfVal,
		int	nStartIdx,
		int	nCount
	)

Set a data range with a given value.

Parameters

dfVal	Specifies the value to set.
nStartIdx	Specifies the start index.
nCount	Specifies the number of items to set.

Definition at line 1961 of file Blob.cs.

◆ SetData() [3/5]

void MyCaffe.common.Blob< T >.SetData	(	SimpleDatum	d,
		bool	bReshape,
		bool	bCopyData = `true`
	)

Sets the Blob values to the data contained within a SimpleDatum.

Parameters

d	Specifies the SimpleDatum.
bReshape	Specifies whether or not to reshape the Blob to match the SimpleDatum.
bCopyData	Optionally, specifies whether or not to transfer the data.

Definition at line 2029 of file Blob.cs.

◆ SetData() [4/5]

void MyCaffe.common.Blob< T >.SetData	(	T	fVal,
		int	nIdx = `-1`
	)

Either sets all of the data items in the Blob to a given value, or alternatively only sets a single indexed item to a given value.

Parameters

fVal	Specifies the value to set.
nIdx	Optionally, specifies the index of the item to set.

Definition at line 1933 of file Blob.cs.

◆ SetData() [5/5]

void MyCaffe.common.Blob< T >.SetData	(	T[]	rgData,
		int	nCount = `-1`,
		bool	bSetCount = `true`
	)

Sets a number of items within the Blob's data.

Parameters

rgData	Specifies the data to set.
nCount	Optionally, specifies a subset count of items to set.
bSetCount	Optionally, specifies whether or not to set the count. The count is always set when re-allocating the buffer.

Definition at line 1922 of file Blob.cs.

◆ SetDiff() [1/3]

void MyCaffe.common.Blob< T >.SetDiff	(	double	dfVal,
		int	nIdx = `-1`
	)

Either sets all of the diff items in the Blob to a given value, or alternatively only sets a single indexed item to a given value.

Parameters

dfVal	Specifies the value to set.
nIdx	Optionally, specifies the index of the item to set.

Definition at line 1981 of file Blob.cs.

◆ SetDiff() [2/3]

void MyCaffe.common.Blob< T >.SetDiff	(	double	dfVal,
		int	nStartIdx,
		int	nCount
	)

Set a diff range with a given value.

Parameters

dfVal	Specifies the value to set.
nStartIdx	Specifies the start index.
nCount	Specifies the number of items to set.

Definition at line 1998 of file Blob.cs.

◆ SetDiff() [3/3]

void MyCaffe.common.Blob< T >.SetDiff	(	T[]	rgDiff,
		int	nCount = `-1`,
		bool	bSetCount = `true`
	)

Sets a number of items within the Blob's diff.

Parameters

rgDiff	Specifies the diff to set.
nCount	Optionally, specifies a subset count of items to set.
bSetCount	Optionally, specifies whether or not to set the count. The count is always set when re-allocating the buffer.

Definition at line 2018 of file Blob.cs.

◆ SetParameter()

void MyCaffe.common.Blob< T >.SetParameter	(	string	strName,
		double	dfVal
	)

Set a blob parameter.

Parameters

strName	Specifies the name of the blob parameter.
dfVal	Specifies the value of the blob parameter.

Definition at line 233 of file Blob.cs.

◆ SetPixel() [1/2]

Tuple< T, T, T > MyCaffe.common.Blob< T >.SetPixel	(	int	nX,
		int	nY,
		byte	R,
		byte	G,
		byte	B,
		TransformationParameter.COLOR_ORDER	order = `TransformationParameter.COLOR_ORDER.RGB`
	)

Set the values of a 3 channel image embedded within the blob in the order RGB or BGR where the full hxw is stored for each color in order.

Parameters

nX	Specifies the x pixel within the image.
nY	Specifies the y pxiel within the image.
R	Specifies the color for the Red color channel.
G	Specifies the color for the Green color channel.
B	Specifies the color for the Blue color channel.
order	Optionally, specifies the color ordering RGB or BGR (default = RGB).

Returns: The original pixel is returned as a three item tuple.

Definition at line 3035 of file Blob.cs.

◆ SetPixel() [2/2]

Tuple< T, T, T > MyCaffe.common.Blob< T >.SetPixel	(	int	nX,
		int	nY,
		Tuple< T, T, T >	pixel,
		bool	bReturnOriginal = `false`,
		TransformationParameter.COLOR_ORDER	order = `TransformationParameter.COLOR_ORDER.RGB`,
		int	nOffset = `0`
	)

Sets a pixel to the values within a three item tuple where the first item is assigned RED, second GREEN and third BLUE.

Parameters

nX	Specifies the x pixel within the image.
nY	Specifies the y pxiel within the image.
pixel	Specifies the color values for RED, GREEN and BLUE.
bReturnOriginal	Optionally, specifies to return the original pixel.
order	Optionally, specifies the color ordering RGB or BGR (default = RGB).
nOffset	Optionally, specifies the offset to start at (default = 0).

Returns: The original pixel is returned as a three item tuple.

Definition at line 3068 of file Blob.cs.

◆ shape() [1/2]

List< int > MyCaffe.common.Blob< T >.shape ( )

Returns an array where each element contains the shape of an axis of the Blob.

Returns: The shape array is returned.

Definition at line 684 of file Blob.cs.

◆ shape() [2/2]

int MyCaffe.common.Blob< T >.shape ( int nIdx )

Returns the dimension of the nIdx'th axis (or the negative nIdx'th axis from teh end, if nIdx is negative.

Parameters

nIdx	The axis index, which may be negative as it will be 'canonicalized' using CanonicalAxisIndex.

Returns

Definition at line 696 of file Blob.cs.

◆ ShapeEquals()

bool MyCaffe.common.Blob< T >.ShapeEquals ( BlobProto bp )

Compares the shape of this blob to the shape within a BlobProto.

Parameters

bp	Specifies the BlobProto to compare.

Returns: If the shapes are the same this method returns true, otherwise false.

Definition at line 2079 of file Blob.cs.

◆ Share()

void MyCaffe.common.Blob< T >.Share ( Blob< T > b )

Share another Blob with this one, by setting the data and diff to the same data and diff of the other Blob.

Parameters

b	Specifies the other Blob to share.

Definition at line 1850 of file Blob.cs.

◆ ShareData()

void MyCaffe.common.Blob< T >.ShareData ( Blob< T > b )

Set the data to point to the data of the other blob – useful in Layers which simply perform a copy in their forward pass.

Parameters

b

Definition at line 1813 of file Blob.cs.

◆ ShareDiff()

void MyCaffe.common.Blob< T >.ShareDiff ( Blob< T > b )

Set the diff to point to the diff of the other blob – useful in Layers which simply perform a copy in their forward pass.

Parameters

b

Definition at line 1832 of file Blob.cs.

◆ std()

double MyCaffe.common.Blob< T >.std	(	double?	dfMean = `null`,
		float[]	rgDf = `null`
	)

Calculate the standard deviation of the blob data.

Parameters

dfMean	Optionally, specifies the mean to use (default = null).
rgDf	Optionally, specifies the CPU data to calculate the std on (default = null).

Returns: The standard deviation of the bob data is returned.

Definition at line 3007 of file Blob.cs.

◆ sum()

double MyCaffe.common.Blob< T >.sum	(	float[]	rgDf = `null`,
		bool	bDiff = `false`
	)

Calculate the sum of the blob data.

Parameters

rgDf	Optionally, specifies the CPU data to use (default = null).
bDiff	Optionally, specifies to use the diff instead of the data.

Returns: The mean is returned.

Definition at line 2986 of file Blob.cs.

◆ sumsq_data()

T MyCaffe.common.Blob< T >.sumsq_data ( )

Calcualte the sum of squares (L2 norm squared) of the data.

Returns

Definition at line 1730 of file Blob.cs.

◆ sumsq_diff()

T MyCaffe.common.Blob< T >.sumsq_diff ( )

Calcualte the sum of squares (L2 norm squared) of the diff.

Returns

Definition at line 1742 of file Blob.cs.

◆ ToByteArray()

byte[] MyCaffe.common.Blob< T >.ToByteArray ( )

Saves this Blob to a byte array.

Returns: The byte array containing all data of the blob is returned.

Definition at line 2436 of file Blob.cs.

◆ ToDatum()

Datum MyCaffe.common.Blob< T >.ToDatum ( )

Returns a new Datum that contains the shape and data of the Blob.

Returns: A Datum is returned.

Definition at line 2166 of file Blob.cs.

◆ ToProto()

BlobProto MyCaffe.common.Blob< T >.ToProto ( bool bWriteDiff = false )

Writes the Blob to a new BlobProto.

Parameters

bWriteDiff When true, the diff is written to the BlobProto, otherwise the data is written.

Returns: The new BlobProto is returned.

Definition at line 1663 of file Blob.cs.

◆ ToSizeString()

string MyCaffe.common.Blob< T >.ToSizeString ( )

Returns a string describing the 4D shape of the Blob.

Returns: A shape string is returned.

Definition at line 2157 of file Blob.cs.

◆ ToString() [1/2]

override string MyCaffe.common.Blob< T >.ToString ( )

Returns a string representation of the Blob.

Returns

Definition at line 2466 of file Blob.cs.

◆ ToString() [2/2]

string MyCaffe.common.Blob< T >.ToString	(	int	nMax,
		bool	bDiff = `false`
	)

Get the string representation containing up to the first 'nMax' items.

Parameters

nMax	Specifies the maximum number of data items to return.
bDiff	Specifies to returive the diff when true.

Returns: The string representation is returned.

Definition at line 2478 of file Blob.cs.

◆ Unsqueeze()

void MyCaffe.common.Blob< T >.Unsqueeze ( int nNumAxes )

Unsqueeze the shape by adding shape=1 on each axis until the 'nNumAxes' is reached.

Parameters

nNumAxes Specifies the number of axes to unsqueeze to.

Definition at line 201 of file Blob.cs.

◆ Update()

void MyCaffe.common.Blob< T >.Update ( )

The 'update' method is used for parameter blobs in a Net.

Update is called to apply the diff errors to the data. When !bIncludeDiff or freeze_learning = true, no diff is applied.

Definition at line 1575 of file Blob.cs.

◆ update_cpu_data()

T[] MyCaffe.common.Blob< T >.update_cpu_data ( )

Update the CPU data by transferring the GPU data over to the Host.

Returns: The Host data is returned.

Definition at line 1470 of file Blob.cs.

◆ update_cpu_diff()

T[] MyCaffe.common.Blob< T >.update_cpu_diff ( )

Update the CPU diff by transferring the GPU diff over to the Host.

Returns: The Host diff is returned.

Definition at line 1529 of file Blob.cs.

◆ ValidateData()

bool MyCaffe.common.Blob< T >.ValidateData	(	Blob< T >	work,
		bool	bDiff = `false`
	)

Validate the data or diff looking for NAN or Inf.

Parameters

work	Specifies the work blob.
bDiff	Specifies to validate the diff (true) or the data (false).

Returns: True is returned if the data (or diff) contains no NAN or Inf.

Definition at line 1197 of file Blob.cs.

Member Data Documentation

◆ MAX_BLOB_AXES

const int MyCaffe.common.Blob< T >.MAX_BLOB_AXES = 32

static

Defines the maximum number of Axes supported by the Blob.

Definition at line 55 of file Blob.cs.

Property Documentation

◆ channels

int MyCaffe.common.Blob< T >.channels

get

DEPRECIATED; legacy shape accessor channels: use shape(1) instead.

Definition at line 799 of file Blob.cs.

◆ cpu_data

T [] MyCaffe.common.Blob< T >.cpu_data

get

Returns the last host data retrieved from the GPU.

Definition at line 1451 of file Blob.cs.

◆ cpu_diff

T [] MyCaffe.common.Blob< T >.cpu_diff

get

Returns the last host diff retrieved from the GPU.

Definition at line 1495 of file Blob.cs.

◆ Cuda

CudaDnn<T> MyCaffe.common.Blob< T >.Cuda

get

Returns the CudaDnn object that manages the Blob's memory."/>

Definition at line 385 of file Blob.cs.

◆ data

SyncedMemory<T> MyCaffe.common.Blob< T >.data

get

Returns the SyncedMemory that stores the data.

Definition at line 1435 of file Blob.cs.

◆ diff

SyncedMemory<T> MyCaffe.common.Blob< T >.diff

get

Returns the SyncedMemory that stores the diff.

Definition at line 1443 of file Blob.cs.

◆ DiffExists

bool MyCaffe.common.Blob< T >.DiffExists

get

Returns whether or not the Diff portion exists.

Definition at line 675 of file Blob.cs.

◆ freeze_learning

bool MyCaffe.common.Blob< T >.freeze_learning

getset

Specifies whether or not the diff is applied to the data during Update. When freeze learning = true, the update is skipped.

Definition at line 376 of file Blob.cs.

◆ gpu_data

long MyCaffe.common.Blob< T >.gpu_data

get

Returns the data GPU handle used by the CudaDnn connection.

Definition at line 1478 of file Blob.cs.

◆ gpu_diff

long MyCaffe.common.Blob< T >.gpu_diff

get

Returns the diff GPU handle used by the CudaDnn connection.

Definition at line 1540 of file Blob.cs.

◆ gpu_shape

long MyCaffe.common.Blob< T >.gpu_shape

get

Returns the shape GPU handle used by the CudaDnn connection. The shape data contains the shape information of the Blob for use in GPU operations.

Definition at line 1564 of file Blob.cs.

◆ HalfSize

bool MyCaffe.common.Blob< T >.HalfSize

get

Returns whether or not this blob is using half sizes.

Definition at line 368 of file Blob.cs.

◆ height

int MyCaffe.common.Blob< T >.height

get

DEPRECIATED; legacy shape accessor height: use shape(2) instead.

Definition at line 807 of file Blob.cs.

◆ Log

Log MyCaffe.common.Blob< T >.Log

get

Returns the Log associated with the blob.

Definition at line 393 of file Blob.cs.

◆ max_data

double MyCaffe.common.Blob< T >.max_data

get

Returns the maximum value in the data of the Blob.

Definition at line 2524 of file Blob.cs.

◆ max_diff

double MyCaffe.common.Blob< T >.max_diff

get

Returns the maximum value in the diff of the Blob.

Definition at line 2576 of file Blob.cs.

◆ min_data

double MyCaffe.common.Blob< T >.min_data

get

Returns the minimum value in the data of the Blob.

Definition at line 2498 of file Blob.cs.

◆ min_diff

double MyCaffe.common.Blob< T >.min_diff

get

Returns the minimum value in the diff of the Blob.

Definition at line 2550 of file Blob.cs.

◆ MinusOne

T MyCaffe.common.Blob< T >.MinusOne

staticget

Returns MinusOne (-1) in type T.

Definition at line 275 of file Blob.cs.

◆ mutable_cpu_data

T [] MyCaffe.common.Blob< T >.mutable_cpu_data

getset

Get data from the GPU and bring it over to the host, or Set data from the Host and send it over to the GPU.

Definition at line 1460 of file Blob.cs.

◆ mutable_cpu_diff

T [] MyCaffe.common.Blob< T >.mutable_cpu_diff

getset

Get diff from the GPU and bring it over to the host, or Set diff from the Host and send it over to the GPU.

Definition at line 1510 of file Blob.cs.

◆ mutable_gpu_data

long MyCaffe.common.Blob< T >.mutable_gpu_data

get

Returns the data GPU handle used by the CudaDnn connection.

Definition at line 1486 of file Blob.cs.

◆ mutable_gpu_diff

long MyCaffe.common.Blob< T >.mutable_gpu_diff

get

Returns the diff GPU handle used by the CudaDnn connection.

Definition at line 1554 of file Blob.cs.

◆ Name

string MyCaffe.common.Blob< T >.Name

getset

Get/set the name of the Blob.

Definition at line 2183 of file Blob.cs.

◆ num

int MyCaffe.common.Blob< T >.num

get

DEPRECIATED; legacy shape accessor num: use shape(0) instead.

Definition at line 791 of file Blob.cs.

◆ num_axes

int MyCaffe.common.Blob< T >.num_axes

get

Returns the number of axes in the Blob.

Definition at line 704 of file Blob.cs.

◆ num_true_axes

int MyCaffe.common.Blob< T >.num_true_axes

get

Returns the number of true axes, ignoring the trailing ones.

Definition at line 715 of file Blob.cs.

◆ One

T MyCaffe.common.Blob< T >.One

staticget

Returns One (1) in type T.

Definition at line 267 of file Blob.cs.

◆ Padded

bool MyCaffe.common.Blob< T >.Padded

getset

Get/set the padding state of the blob.

Definition at line 283 of file Blob.cs.

◆ reshape_when_sharing

bool MyCaffe.common.Blob< T >.reshape_when_sharing

getset

When true, this Blob is reshaped to the source when sharing the source data (default = false).

This setting is used by the Net when sharing trained weights.

Definition at line 1802 of file Blob.cs.

◆ shape_string

string MyCaffe.common.Blob< T >.shape_string

get

Returns a string describing the Blob's shape.

Definition at line 656 of file Blob.cs.

◆ snapshot_requested

bool MyCaffe.common.Blob< T >.snapshot_requested

getset

Get/set the snapshot request.

This setting is used by learnable parameters that have requested a snapshot.

Definition at line 1882 of file Blob.cs.

◆ Tag

object MyCaffe.common.Blob< T >.Tag

getset

Returns a user defined object associated with the Blob.

Definition at line 2769 of file Blob.cs.

◆ type

BLOB_TYPE MyCaffe.common.Blob< T >.type

getset

Returns the BLOB_TYPE of the Blob.

Definition at line 2760 of file Blob.cs.

◆ width

int MyCaffe.common.Blob< T >.width

get

DEPRECIATED; legacy shape accessor width: use shape(3) instead.

Definition at line 815 of file Blob.cs.

◆ Zero

T MyCaffe.common.Blob< T >.Zero

staticget

Returns Zero (0) in type T.

Definition at line 259 of file Blob.cs.

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

C:/Data/Data/SS_Projects/Intelligence/GitHub/MyCaffe/MyCaffe/common/Blob.cs

Public Member Functions

Static Public Member Functions

Static Public Attributes

Protected Member Functions

Properties

Detailed Description

Constructor & Destructor Documentation

◆ Blob() [1/8]

◆ Blob() [2/8]

◆ Blob() [3/8]

◆ Blob() [4/8]

◆ Blob() [5/8]

◆ Blob() [6/8]

◆ Blob() [7/8]

◆ Blob() [8/8]

Member Function Documentation

◆ add_scalar()

◆ asum_data()

◆ asum_diff()

◆ AsyncGpuPush()

◆ CanonicalAxisIndex()

◆ Clone()

◆ Compare() [1/2]

◆ Compare() [2/2]

◆ CompareEx()

◆ CompareShape() [1/2]

◆ CompareShape() [2/2]

◆ ConvertToBase()

◆ ConvertToHalf()

◆ CopyFrom() [1/3]

◆ CopyFrom() [2/3]

◆ CopyFrom() [3/3]

◆ CopyFromAndPad()

◆ CopyFromAndTransposeHeightWidth()

◆ CopyParameters()

◆ count() [1/3]

◆ count() [2/3]

◆ count() [3/3]

◆ data_at() [1/2]

◆ data_at() [2/2]

◆ diff_at() [1/2]

◆ diff_at() [2/2]

◆ Dispose() [1/2]

◆ Dispose() [2/2]

◆ FromByteArray()

◆ FromProto()

◆ GetConversionWorkSize()

◆ GetData()

◆ GetDiff()

◆ GetMaxData()

◆ GetMaxDiff()

◆ GetMinData()

◆ GetMinDiff()

◆ GetParameter()

◆ LegacyShape()

◆ Load()

◆ LoadBinary()

◆ LoadFromNumpy() [1/2]

◆ LoadFromNumpy() [2/2]

◆ MathAdd()

◆ MathDiv()

◆ MathSub()

◆ MatMul()

◆ MatMulGrad()

◆ mean()

◆ minmax_data()

◆ minmax_diff()

◆ NormalizeData()

◆ offset() [1/2]

◆ offset() [2/2]

◆ Percentile()

◆ Reshape() [1/4]

◆ Reshape() [2/4]

◆ Reshape() [3/4]

◆ Reshape() [4/4]

◆ ReshapeLike()

◆ Resize()

◆ Save()

◆ SaveBinary()

◆ SaveToImage()