kaldi.cudamatrix¶

Functions

`approx_equal_cu_double_matrix`	Calls C++ function
`approx_equal_cu_double_vector`	Calls C++ function
`approx_equal_cu_matrix`	Calls C++ function
`approx_equal_cu_vector`	Calls C++ function
`assert_equal_cu_double_matrix`	Calls C++ function
`assert_equal_cu_double_vector`	Calls C++ function
`assert_equal_cu_matrix`	Calls C++ function
`assert_equal_cu_vector`	Calls C++ function
`cuda_available`	Check if cuda is available.
`same_dim_and_stride_cu_double_matrix`	Calls C++ function
`same_dim_and_stride_cu_matrix`	Calls C++ function
`same_dim_cu_double_matrix`	Calls C++ function
`same_dim_cu_matrix`	Calls C++ function
`synchronize_gpu`	Synchronizes threads.
`trace_mat_smat`	If trans==kNoTrans compute Trace(A*B).

Classes

`CuArray`	CLIF wrapper for ::kaldi::CuArray<int>
`CuBlockMatrixData`	CLIF wrapper for ::CuBlockMatrixData_
`CuDevice`	CUDA Device.
`CuDoubleMatrix`	CLIF wrapper for ::kaldi::CuMatrix<double>
`CuDoubleMatrixBase`	CLIF wrapper for ::kaldi::CuMatrixBase<double>
`CuDoublePtrArray`	CLIF wrapper for ::kaldi::CuArray<double *>
`CuDoubleSubMatrix`	CLIF wrapper for ::kaldi::CuSubMatrix<double>
`CuDoubleSubVector`	CLIF wrapper for ::kaldi::CuSubVector<double>
`CuDoubleVector`	CLIF wrapper for ::kaldi::CuVector<double>
`CuDoubleVectorBase`	CLIF wrapper for ::kaldi::CuVectorBase<double>
`CuFloatPtrArray`	CLIF wrapper for ::kaldi::CuArray<float *>
`CuIntPairArray`	CLIF wrapper for ::kaldi::CuArray< ::Int32Pair>
`CuMatrix`	CLIF wrapper for ::kaldi::CuMatrix<float>
`CuMatrixBase`	CLIF wrapper for ::kaldi::CuMatrixBase<float>
`CuSparseMatrix`	This class defines a Sparse CUDA Matrix.
`CuSubMatrix`	CLIF wrapper for ::kaldi::CuSubMatrix<float>
`CuSubVector`	CLIF wrapper for ::kaldi::CuSubVector<float>
`CuTimer`	Convenience wrapper for Timer.
`CuVector`	CLIF wrapper for ::kaldi::CuVector<float>
`CuVectorBase`	CLIF wrapper for ::kaldi::CuVectorBase<float>
`DoubleMatrixElement`	CLIF wrapper for ::MatrixElement<double>
`IntPair`	CLIF wrapper for ::Int32Pair
`MatrixDim`	CLIF wrapper for ::MatrixDim_
`MatrixElement`	CLIF wrapper for ::MatrixElement<float>

class kaldi.cudamatrix.CuArray¶

CLIF wrapper for ::kaldi::CuArray<int>

add(value:int)¶: Add a constant value

copy_from_array(src:CuArray)¶: This function resizes if needed.

copy_from_vec(src:list<int>)¶: This function resizes if needed.

copy_to_vec() → list<int>¶: This function resizes dst if needed.

destroy()¶: Deallocate the memory and set dim_ and data_ to zero.

dim() → int¶

Returns:	This vector dimension

from_other(src:CuArray) → CuArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray<int>> ::kaldi::CuArray<int>::CuArray(::kaldi::CuArray<int>)

from_size(dim:int, resize_type:MatrixResizeType=default) → CuArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray<int>> ::kaldi::CuArray<int>::CuArray(int, ::kaldi::MatrixResizeType)

from_vector(src:list<int>) → CuArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray<int>> ::kaldi::CuArray<int>::CuArray(::std::vector<int>)

max() → int¶

Returns:	Maximum value

min() → int¶

Returns:	Minimum value

read(is:istream, binary:bool)¶: I/O

resize(dim:int, resize_type:MatrixResizeType=default)¶: Allocate the memory.

set(value:int)¶: Set to a constant value.

set_zero()¶: Sets the memory for the object to zero.

write(os:ostream, binary:bool)¶: I/O

class kaldi.cudamatrix.CuBlockMatrixData¶

CLIF wrapper for ::CuBlockMatrixData_

col_offset¶: C++ ::int32_cuda CuBlockMatrixData_.col_offset

matrix_dim¶: C++ ::MatrixDim CuBlockMatrixData_.matrix_dim

row_offset¶: C++ ::int32_cuda CuBlockMatrixData_.row_offset

class kaldi.cudamatrix.CuDevice¶

CUDA Device.

This class is for initializing, selecting and querying a CUDA device.

There is a separate instance of the CuDevice object for each thread of the program, but many of its variables are static (hence, shared between all instances).

allow_multithreading()¶

Enables multi-threaded access to GPU.

The user should call this if the program plans to access the GPU (e.g. via using class CuMatrix) from more than one thread. If you fail to call this for a multi-threaded program, it may occasionally segfault (and also the code will detect that you failed to call it, and will print a warning).

check_gpu_health()¶

Check if GPU is in good condition.

This is done by multiplying small matrices on GPU+CPU. Overheated GPUs may give inaccurate results, which we want to detect.

double_precision_supported() → bool¶

Checks if double precision is supported.

Returns true if either we have no GPU, or we have a GPU and it supports double precision.

enabled() → bool¶: Checks if the CUDA GPU is selected for use.

instantiate() → CuDevice¶

Instantiates a new CuDevice.

You obtain the CuDevice for the current thread by calling CuDevice.instantiate At the beginning of the program, if you want to use a GPU, you should call CuDevice.instantiate().select_gpu_id(..).

is_compute_exclusive() → bool¶

Checks if the GPU is set to compute exclusive mode.

You can set this mode, if you are root, by doing: nvidia-smi -c 3.

Returns True if we have a GPU and it is running in compute exclusive mode. Returns false otherwise. WILL CRASH if we are not using a GPU at all. If calling this as a user (i.e. from outside the class), call this only if enabled() returns True.

print_memory_usage()¶: Print some memory-usage information using KALDI_LOG.

select_gpu_id(use_gpu:str)¶

Selects a GPU for computation.

This function is supposed to be called just once, at the beginning of the program (from the main thread), or not at all.

Parameters: use_gpu (str) – GPU use mode. One of “yes”, “optional” or “no”. “yes” – Select GPU automatically and die if this fails. If you have set the GPUs to exclusive mode it will select one pseudo-randomly; otherwise it will choose whichever one has the most free memory (but we recommend to set GPUs to exclusive mode, or controlling which GPU to use by setting the variable CUDA_VISIBLE_DEVICES to the id of the GPU you want the program to use. “optional” – Do as above, but if it fails, back off to CPU. “no” – Run on CPU.

set_debug_stride_mode(mode:bool) → bool¶

Activates debug stride mode.

Activates a mode where calls to MallocPitch will purposely allocate arrays with different pitch (inconsistent between calls). This is only useful for testing code. This function returns the previous mode, where true means inconsistent pitch. Note that you cannot ever rely on the strides from MallocPitch() being consistent for the same request, but in practice they tend to be consistent unless you are close to running out of memory.

class kaldi.cudamatrix.CuDoubleMatrix¶

CLIF wrapper for ::kaldi::CuMatrix<double>

add(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Add(double)

add_cols(src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddCols(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

add_diag_vec_mat(alpha:float, v:CuDoubleVectorBase, M:CuDoubleMatrixBase, transM:MatrixTransposeType, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddDiagVecMat(double, ::kaldi::CuVectorBase<double>, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, double)

add_elements(alpha:float, input:list<DoubleMatrixElement>)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddElements(double, ::std::vector< ::MatrixElement<double> >)

add_mat(alpha:float, A:CuDoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMat(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType)

add_mat_blocks(alpha:float, A:CuDoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMatBlocks(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType)

add_mat_diag_vec(alpha:float, M:CuDoubleMatrixBase, transM:MatrixTransposeType, v:CuDoubleVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMatDiagVec(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, ::kaldi::CuVectorBase<double>, double)

add_mat_mat(alpha:float, A:CuDoubleMatrixBase, transA:MatrixTransposeType, B:CuDoubleMatrixBase, transB:MatrixTransposeType, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMatMat(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, double)

add_mat_mat_elements(alpha:float, A:CuDoubleMatrixBase, B:CuDoubleMatrixBase, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMatMatElements(double, ::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, double)

add_row_ranges(src:CuDoubleMatrixBase, indexes:CuIntPairArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddRowRanges(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray< ::Int32Pair>)

add_rows(alpha:float, src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddRows(double, ::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

add_to_diag(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddToDiag(double)

add_to_rows(alpha:float, dst:CuDoublePtrArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddToRows(double, ::kaldi::CuArray<double *>)

add_vec_to_cols(alpha:float, col:CuDoubleVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddVecToCols(double, ::kaldi::CuVectorBase<double>, double)

add_vec_to_rows(alpha:float, row:CuDoubleVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddVecToRows(double, ::kaldi::CuVectorBase<double>, double)

add_vec_vec(alpha:float, x:CuDoubleVectorBase, y:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddVecVec(double, ::kaldi::CuVectorBase<double>, ::kaldi::CuVectorBase<double>)

apply_ceiling(ceiling_val:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyCeiling(double)

apply_exp()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyExp()

apply_floor(floor_val:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyFloor(double)

apply_heaviside()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyHeaviside()

apply_log()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyLog()

apply_log_soft_max_per_row(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyLogSoftMaxPerRow(::kaldi::CuMatrixBase<double>)

apply_pow(power:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyPow(double)

apply_pow_abs(power:float, include_sign:bool=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyPowAbs(double, bool)

apply_soft_max_per_row(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplySoftMaxPerRow(::kaldi::CuMatrixBase<double>)

approx_equal(other:CuDoubleMatrixBase, tol:float=default) → bool¶: Calls C++ function bool ::kaldi::CuMatrixBase<double>::ApproxEqual(::kaldi::CuMatrixBase<double>, float)

cholesky(inv_cholesky:CuDoubleMatrixBase=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Cholesky(::kaldi::CuMatrixBase<double> *)

comp_objf_and_deriv(elements:list<DoubleMatrixElement>, A:CuDoubleMatrix) -> (tot_objf:float, tot_weight:float)¶: Calls C++ function void ::kaldi::CuMatrix<double>::CompObjfAndDeriv(::std::vector< ::MatrixElement<double> >, ::kaldi::CuMatrix<double>, double*, double*)

copy_col_from_cu_vec(v:CuDoubleVectorBase, col:int)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyColFromVec(::kaldi::CuVectorBase<double>, int)

copy_cols(src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyCols(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

copy_cols_from_cu_vec(v:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyColsFromVec(::kaldi::CuVectorBase<double>)

copy_from_cu_mat(src:CuDoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyFromMat(::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType)

copy_from_mat(src:DoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyFromMat(::kaldi::MatrixBase<double>, ::kaldi::MatrixTransposeType)

copy_lower_to_upper()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyLowerToUpper()

copy_rows(src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyRows(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

copy_rows_from_cu_vec(v:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyRowsFromVec(::kaldi::CuVectorBase<double>)

copy_rows_from_vec(v:DoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyRowsFromVec(::kaldi::VectorBase<double>)

copy_to_mat(dst:DoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyToMat(::kaldi::MatrixBase<double> *, ::kaldi::MatrixTransposeType)

copy_to_rows(dst:CuDoublePtrArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyToRows(::kaldi::CuArray<double *>)

copy_upper_to_lower()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyUpperToLower()

diff_group_pnorm(in_value:CuDoubleMatrixBase, out_value:CuDoubleMatrixBase, out_deriv:CuDoubleMatrixBase, power:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffGroupPnorm(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, double)

diff_log_softmax_per_row(out_value:CuDoubleMatrixBase, out_deriv:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffLogSoftmaxPerRow(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

diff_parametric_relu(value:CuDoubleMatrixBase, diff:CuDoubleMatrixBase, alpha:CuDoubleVectorBase, beta:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffParametricRelu(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, ::kaldi::CuVectorBase<double>, ::kaldi::CuVectorBase<double>)

diff_sigmoid(value:CuDoubleMatrixBase, diff:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffSigmoid(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

diff_softmax_per_row(value:CuDoubleMatrixBase, diff:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffSoftmaxPerRow(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

diff_tanh(value:CuDoubleMatrixBase, diff:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffTanh(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

diff_xent(tgt:CuArray) → CuDoubleVector¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffXent(::kaldi::CuArray<int>, ::kaldi::CuVector<double>*)

dim() → MatrixDim¶: Calls C++ function ::MatrixDim_ ::kaldi::CuMatrixBase<double>::Dim()

div_elements(A:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DivElements(::kaldi::CuMatrixBase<double>)

div_rows_vec(div:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DivRowsVec(::kaldi::CuVectorBase<double>)

find_row_max_id() → CuArray¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::FindRowMaxId(::kaldi::CuArray<int>*)

frobenius_norm() → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::FrobeniusNorm()

from_matrix(other:DoubleMatrixBase, trans:MatrixTransposeType=default) → CuDoubleMatrix¶: Calls C++ function std::unique_ptr<::kaldi::CuMatrix<double>> ::kaldi::CuMatrix<double>::CuMatrix(::kaldi::MatrixBase<double>, ::kaldi::MatrixTransposeType)

from_other(other:CuDoubleMatrixBase, trans:MatrixTransposeType=default) → CuDoubleMatrix¶: Calls C++ function std::unique_ptr<::kaldi::CuMatrix<double>> ::kaldi::CuMatrix<double>::CuMatrix(::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType)

from_size(rows:int, cols:int, resize_type:MatrixResizeType=default, stride_type:MatrixStrideType=default) → CuDoubleMatrix¶: Calls C++ function std::unique_ptr<::kaldi::CuMatrix<double>> ::kaldi::CuMatrix<double>::CuMatrix(int, int, ::kaldi::MatrixResizeType, ::kaldi::MatrixStrideType)

group_max(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::GroupMax(::kaldi::CuMatrixBase<double>)

group_max_deriv(input:CuDoubleMatrixBase, output:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::GroupMaxDeriv(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

group_pnorm(src:CuDoubleMatrixBase, pow:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::GroupPnorm(::kaldi::CuMatrixBase<double>, double)

heaviside(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Heaviside(::kaldi::CuMatrixBase<double>)

invert_elements()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::InvertElements()

is_unit(tol:float=default) → bool¶: Calls C++ function bool ::kaldi::CuMatrixBase<double>::IsUnit(double)

max() → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::Max()

max_elementwise(A:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Max(::kaldi::CuMatrixBase<double>)

min() → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::Min()

min_elementwise(A:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Min(::kaldi::CuMatrixBase<double>)

mul_cols_vec(scale:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulColsVec(::kaldi::CuVectorBase<double>)

mul_elements(A:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulElements(::kaldi::CuMatrixBase<double>)

mul_rows(src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulRows(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

mul_rows_group_mat(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulRowsGroupMat(::kaldi::CuMatrixBase<double>)

mul_rows_vec(scale:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulRowsVec(::kaldi::CuVectorBase<double>)

num_cols() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<double>::NumCols()

num_rows() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<double>::NumRows()

parametric_relu(src:CuDoubleMatrixBase, alpha:CuDoubleVectorBase, beta:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ParametricRelu(::kaldi::CuMatrixBase<double>, ::kaldi::CuVectorBase<double>, ::kaldi::CuVectorBase<double>)

read(is:istream, binary:bool)¶: Calls C++ function void ::kaldi::CuMatrix<double>::Read(::std::basic_istream<char, ::std::char_traits<char> >, bool)

resize(rows:int, cols:int, resize_type:MatrixResizeType=default, stride_type:MatrixStrideType=default)¶: Calls C++ function void ::kaldi::CuMatrix<double>::Resize(int, int, ::kaldi::MatrixResizeType, ::kaldi::MatrixStrideType)

scale(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Scale(double)

set(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Set(double)

set_mat_mat_div_mat(A:CuDoubleMatrixBase, B:CuDoubleMatrixBase, C:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetMatMatDivMat(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

set_rand_uniform()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetRandUniform()

set_randn()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetRandn()

set_zero()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetZero()

set_zero_above_diag()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetZeroAboveDiag()

sigmoid(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Sigmoid(::kaldi::CuMatrixBase<double>)

size_in_bytes() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<double>::SizeInBytes()

soft_hinge(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SoftHinge(::kaldi::CuMatrixBase<double>)

stride() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<double>::Stride()

sum() → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::Sum()

sum_column_ranges(src:CuDoubleMatrixBase, indexes:CuIntPairArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SumColumnRanges(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray< ::Int32Pair>)

swap(mat:CuDoubleMatrix)¶: Calls C++ function void ::kaldi::CuMatrix<double>::Swap(::kaldi::CuMatrix<double> *)

swap_with_matrix(mat:DoubleMatrix)¶: Calls C++ function void ::kaldi::CuMatrix<double>::Swap(::kaldi::Matrix<double> *)

sym_add_mat2(alpha:float, M:CuDoubleMatrixBase, transA:MatrixTransposeType, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SymAddMat2(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, double)

sym_invert_pos_def()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SymInvertPosDef()

tanh(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Tanh(::kaldi::CuMatrixBase<double>)

trace(check_square:bool=default) → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::Trace(bool)

transpose()¶: Calls C++ function void ::kaldi::CuMatrix<double>::Transpose()

write(os:ostream, binary:bool)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Write(::std::basic_ostream<char, ::std::char_traits<char> >, bool)

class kaldi.cudamatrix.CuDoublePtrArray¶

CLIF wrapper for ::kaldi::CuArray<double *>

copy_from_array(src:CuDoublePtrArray)¶: This function resizes if needed.

destroy()¶: Deallocate the memory and set dim_ and data_ to zero.

dim() → int¶

Returns:	This vector dimension

from_other(src:CuDoublePtrArray) → CuDoublePtrArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray<double *>> ::kaldi::CuArray<double *>::CuArray(::kaldi::CuArray<double *>)

from_size(dim:int, resize_type:MatrixResizeType=default) → CuDoublePtrArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray<double *>> ::kaldi::CuArray<double *>::CuArray(int, ::kaldi::MatrixResizeType)

max() → float¶

Returns:	Maximum value

min() → float¶

Returns:	Minimum value

resize(dim:int, resize_type:MatrixResizeType=default)¶: Allocate the memory.

set_zero()¶: Sets the memory for the object to zero.

class kaldi.cudamatrix.CuDoubleSubMatrix¶

CLIF wrapper for ::kaldi::CuSubMatrix<double>

add(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Add(double)

add_cols(src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddCols(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

add_diag_vec_mat(alpha:float, v:CuDoubleVectorBase, M:CuDoubleMatrixBase, transM:MatrixTransposeType, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddDiagVecMat(double, ::kaldi::CuVectorBase<double>, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, double)

add_elements(alpha:float, input:list<DoubleMatrixElement>)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddElements(double, ::std::vector< ::MatrixElement<double> >)

add_mat(alpha:float, A:CuDoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMat(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType)

add_mat_blocks(alpha:float, A:CuDoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMatBlocks(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType)

add_mat_diag_vec(alpha:float, M:CuDoubleMatrixBase, transM:MatrixTransposeType, v:CuDoubleVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMatDiagVec(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, ::kaldi::CuVectorBase<double>, double)

add_mat_mat(alpha:float, A:CuDoubleMatrixBase, transA:MatrixTransposeType, B:CuDoubleMatrixBase, transB:MatrixTransposeType, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMatMat(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, double)

add_mat_mat_elements(alpha:float, A:CuDoubleMatrixBase, B:CuDoubleMatrixBase, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddMatMatElements(double, ::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, double)

add_row_ranges(src:CuDoubleMatrixBase, indexes:CuIntPairArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddRowRanges(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray< ::Int32Pair>)

add_rows(alpha:float, src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddRows(double, ::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

add_to_diag(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddToDiag(double)

add_to_rows(alpha:float, dst:CuDoublePtrArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddToRows(double, ::kaldi::CuArray<double *>)

add_vec_to_cols(alpha:float, col:CuDoubleVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddVecToCols(double, ::kaldi::CuVectorBase<double>, double)

add_vec_to_rows(alpha:float, row:CuDoubleVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddVecToRows(double, ::kaldi::CuVectorBase<double>, double)

add_vec_vec(alpha:float, x:CuDoubleVectorBase, y:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::AddVecVec(double, ::kaldi::CuVectorBase<double>, ::kaldi::CuVectorBase<double>)

apply_ceiling(ceiling_val:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyCeiling(double)

apply_exp()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyExp()

apply_floor(floor_val:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyFloor(double)

apply_heaviside()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyHeaviside()

apply_log()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyLog()

apply_log_soft_max_per_row(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyLogSoftMaxPerRow(::kaldi::CuMatrixBase<double>)

apply_pow(power:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyPow(double)

apply_pow_abs(power:float, include_sign:bool=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplyPowAbs(double, bool)

apply_soft_max_per_row(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ApplySoftMaxPerRow(::kaldi::CuMatrixBase<double>)

approx_equal(other:CuDoubleMatrixBase, tol:float=default) → bool¶: Calls C++ function bool ::kaldi::CuMatrixBase<double>::ApproxEqual(::kaldi::CuMatrixBase<double>, float)

cholesky(inv_cholesky:CuDoubleMatrixBase=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Cholesky(::kaldi::CuMatrixBase<double> *)

copy_col_from_cu_vec(v:CuDoubleVectorBase, col:int)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyColFromVec(::kaldi::CuVectorBase<double>, int)

copy_cols(src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyCols(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

copy_cols_from_cu_vec(v:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyColsFromVec(::kaldi::CuVectorBase<double>)

copy_from_cu_mat(src:CuDoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyFromMat(::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType)

copy_from_mat(src:DoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyFromMat(::kaldi::MatrixBase<double>, ::kaldi::MatrixTransposeType)

copy_lower_to_upper()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyLowerToUpper()

copy_rows(src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyRows(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

copy_rows_from_cu_vec(v:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyRowsFromVec(::kaldi::CuVectorBase<double>)

copy_rows_from_vec(v:DoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyRowsFromVec(::kaldi::VectorBase<double>)

copy_to_mat(dst:DoubleMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyToMat(::kaldi::MatrixBase<double> *, ::kaldi::MatrixTransposeType)

copy_to_rows(dst:CuDoublePtrArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyToRows(::kaldi::CuArray<double *>)

copy_upper_to_lower()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::CopyUpperToLower()

diff_group_pnorm(in_value:CuDoubleMatrixBase, out_value:CuDoubleMatrixBase, out_deriv:CuDoubleMatrixBase, power:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffGroupPnorm(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, double)

diff_log_softmax_per_row(out_value:CuDoubleMatrixBase, out_deriv:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffLogSoftmaxPerRow(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

diff_parametric_relu(value:CuDoubleMatrixBase, diff:CuDoubleMatrixBase, alpha:CuDoubleVectorBase, beta:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffParametricRelu(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, ::kaldi::CuVectorBase<double>, ::kaldi::CuVectorBase<double>)

diff_sigmoid(value:CuDoubleMatrixBase, diff:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffSigmoid(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

diff_softmax_per_row(value:CuDoubleMatrixBase, diff:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffSoftmaxPerRow(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

diff_tanh(value:CuDoubleMatrixBase, diff:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffTanh(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

diff_xent(tgt:CuArray) → CuDoubleVector¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DiffXent(::kaldi::CuArray<int>, ::kaldi::CuVector<double>*)

dim() → MatrixDim¶: Calls C++ function ::MatrixDim_ ::kaldi::CuMatrixBase<double>::Dim()

div_elements(A:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DivElements(::kaldi::CuMatrixBase<double>)

div_rows_vec(div:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::DivRowsVec(::kaldi::CuVectorBase<double>)

find_row_max_id() → CuArray¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::FindRowMaxId(::kaldi::CuArray<int>*)

frobenius_norm() → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::FrobeniusNorm()

from_other(other:CuDoubleSubMatrix) → CuDoubleSubMatrix¶: Calls C++ function std::unique_ptr<::kaldi::CuSubMatrix<double>> ::kaldi::CuSubMatrix<double>::CuSubMatrix(::kaldi::CuSubMatrix<double>)

group_max(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::GroupMax(::kaldi::CuMatrixBase<double>)

group_max_deriv(input:CuDoubleMatrixBase, output:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::GroupMaxDeriv(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

group_pnorm(src:CuDoubleMatrixBase, pow:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::GroupPnorm(::kaldi::CuMatrixBase<double>, double)

heaviside(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Heaviside(::kaldi::CuMatrixBase<double>)

invert_elements()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::InvertElements()

is_unit(tol:float=default) → bool¶: Calls C++ function bool ::kaldi::CuMatrixBase<double>::IsUnit(double)

max() → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::Max()

max_elementwise(A:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Max(::kaldi::CuMatrixBase<double>)

min() → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::Min()

min_elementwise(A:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Min(::kaldi::CuMatrixBase<double>)

mul_cols_vec(scale:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulColsVec(::kaldi::CuVectorBase<double>)

mul_elements(A:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulElements(::kaldi::CuMatrixBase<double>)

mul_rows(src:CuDoubleMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulRows(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray<int>)

mul_rows_group_mat(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulRowsGroupMat(::kaldi::CuMatrixBase<double>)

mul_rows_vec(scale:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::MulRowsVec(::kaldi::CuVectorBase<double>)

num_cols() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<double>::NumCols()

num_rows() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<double>::NumRows()

parametric_relu(src:CuDoubleMatrixBase, alpha:CuDoubleVectorBase, beta:CuDoubleVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::ParametricRelu(::kaldi::CuMatrixBase<double>, ::kaldi::CuVectorBase<double>, ::kaldi::CuVectorBase<double>)

scale(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Scale(double)

set(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Set(double)

set_mat_mat_div_mat(A:CuDoubleMatrixBase, B:CuDoubleMatrixBase, C:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetMatMatDivMat(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

set_rand_uniform()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetRandUniform()

set_randn()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetRandn()

set_zero()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetZero()

set_zero_above_diag()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SetZeroAboveDiag()

sigmoid(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Sigmoid(::kaldi::CuMatrixBase<double>)

size_in_bytes() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<double>::SizeInBytes()

soft_hinge(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SoftHinge(::kaldi::CuMatrixBase<double>)

stride() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<double>::Stride()

sum() → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::Sum()

sum_column_ranges(src:CuDoubleMatrixBase, indexes:CuIntPairArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SumColumnRanges(::kaldi::CuMatrixBase<double>, ::kaldi::CuArray< ::Int32Pair>)

sym_add_mat2(alpha:float, M:CuDoubleMatrixBase, transA:MatrixTransposeType, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SymAddMat2(double, ::kaldi::CuMatrixBase<double>, ::kaldi::MatrixTransposeType, double)

sym_invert_pos_def()¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::SymInvertPosDef()

tanh(src:CuDoubleMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Tanh(::kaldi::CuMatrixBase<double>)

trace(check_square:bool=default) → float¶: Calls C++ function double ::kaldi::CuMatrixBase<double>::Trace(bool)

write(os:ostream, binary:bool)¶: Calls C++ function void ::kaldi::CuMatrixBase<double>::Write(::std::basic_ostream<char, ::std::char_traits<char> >, bool)

class kaldi.cudamatrix.CuDoubleSubVector¶

CLIF wrapper for ::kaldi::CuSubVector<double>

add(value:float)¶: Adds a constant to this vector

add_vec(alpha:float, vec:CuDoubleVectorBase, beta:float=default)¶: Computes this <– beta * this + alpha * vec

add_vec_vec(alpha:float, v:CuDoubleVectorBase, r:CuDoubleVectorBase, beta:float)¶: Computes this <– alpha * v + beta * r

apply_ceiling(ceiling_val:float) → int¶

Applies ceiling to this vector’s elements

Returns:	The number of elements set to ceiling value.

apply_ceiling_no_count(ceiling_val:float)¶: Applies ceiling to this vector’s elements

apply_exp()¶: Applies exp to this vector’s elements

apply_floor(floor_val:float) → int¶

Applies floor to this vector’s elements

Returns:	The number of elements set to floor value.

apply_floor_no_count(floor_val:float)¶: Applies floor to this vector’s elements

apply_log()¶: Applies log to this vector’s elements

apply_pow(power:float)¶: Applies pow to this vector’s elements

apply_softmax()¶: Applies a softmax to this vector’s elements

approx_equal(other:CuDoubleVectorBase, tol:float=default) → bool¶: Returns true if ||this - other|| <= tol

copy_from_cu_vec(src:CuDoubleVectorBase)¶: Copies data from other CuVector

copy_from_vec(src:DoubleVectorBase)¶: Copies data from other :class:kaldi.matrix.VectorBase

copy_rows_from_mat(M:DoubleMatrixBase)¶: Copies data from :class:kaldi.matrix.MatrixBase

copy_to_vec(dst:DoubleVectorBase)¶: Copies data to :class:kaldi.matrix.VectorBase

dim() → int¶: Returns this vector dimension

from_other(other:CuDoubleSubVector) → CuDoubleSubVector¶: Calls C++ function std::unique_ptr<::kaldi::CuSubVector<double>> ::kaldi::CuSubVector<double>::CuSubVector(::kaldi::CuSubVector<double>)

invert_elements()¶: Inverts this vector’s elements

max() → float¶

Returns:	Maximum value of this vector

min() → float¶

Returns:	Minimum value of this vector

mul_elements(v:CuDoubleVectorBase)¶: Multipy element-by-element by another vector.

norm(p:float) → float¶

Returns:	p-norm of this vector

replace_value(orig:float, changed:float)¶: Set each element to y = (x == orig ? changed : x).

scale(value:float)¶: Multiplies this vector’s data by a constant

set(value:float)¶: Sets this vector’s data to a constant

set_rand_uniform()¶: Sets this vector’s elements to values from random uniform

set_randn()¶: Sets this vector’s elements to random [0,1)

set_zero()¶: Sets this vector’s data to zero

sum() → float¶: Sum this vector’s elements

class kaldi.cudamatrix.CuDoubleVector¶

CLIF wrapper for ::kaldi::CuVector<double>

add(value:float)¶: Adds a constant to this vector

add_vec(alpha:float, vec:CuDoubleVectorBase, beta:float=default)¶: Computes this <– beta * this + alpha * vec

add_vec_vec(alpha:float, v:CuDoubleVectorBase, r:CuDoubleVectorBase, beta:float)¶: Computes this <– alpha * v + beta * r

apply_ceiling(ceiling_val:float) → int¶

Applies ceiling to this vector’s elements

Returns:	The number of elements set to ceiling value.

apply_ceiling_no_count(ceiling_val:float)¶: Applies ceiling to this vector’s elements

apply_exp()¶: Applies exp to this vector’s elements

apply_floor(floor_val:float) → int¶

Applies floor to this vector’s elements

Returns:	The number of elements set to floor value.

apply_floor_no_count(floor_val:float)¶: Applies floor to this vector’s elements

apply_log()¶: Applies log to this vector’s elements

apply_pow(power:float)¶: Applies pow to this vector’s elements

apply_softmax()¶: Applies a softmax to this vector’s elements

approx_equal(other:CuDoubleVectorBase, tol:float=default) → bool¶: Returns true if ||this - other|| <= tol

copy_from_cu_vec(src:CuDoubleVectorBase)¶: Copies data from other CuVector

copy_from_vec(src:DoubleVectorBase)¶: Copies data from other :class:kaldi.matrix.VectorBase

copy_rows_from_mat(M:DoubleMatrixBase)¶: Copies data from :class:kaldi.matrix.MatrixBase

copy_to_vec(dst:DoubleVectorBase)¶: Copies data to :class:kaldi.matrix.VectorBase

dim() → int¶: Returns this vector dimension

from_base(v:CuDoubleVectorBase) → CuDoubleVector¶: Calls C++ function std::unique_ptr<::kaldi::CuVector<double>> ::kaldi::CuVector<double>::CuVector(::kaldi::CuVectorBase<double>)

from_size(dim:int, resize_type:MatrixResizeType=default) → CuDoubleVector¶: Calls C++ function std::unique_ptr<::kaldi::CuVector<double>> ::kaldi::CuVector<double>::CuVector(int, ::kaldi::MatrixResizeType)

from_vector(v:DoubleVectorBase) → CuDoubleVector¶: Calls C++ function std::unique_ptr<::kaldi::CuVector<double>> ::kaldi::CuVector<double>::CuVector(::kaldi::VectorBase<double>)

invert_elements()¶: Inverts this vector’s elements

max() → float¶

Returns:	Maximum value of this vector

min() → float¶

Returns:	Minimum value of this vector

mul_elements(v:CuDoubleVectorBase)¶: Multipy element-by-element by another vector.

norm(p:float) → float¶

Returns:	p-norm of this vector

read(is:istream, binary:bool)¶: I/O

replace_value(orig:float, changed:float)¶: Set each element to y = (x == orig ? changed : x).

resize(dim:int, resize_type:MatrixResizeType=default)¶: Resize this vector to dim

scale(value:float)¶: Multiplies this vector’s data by a constant

set(value:float)¶: Sets this vector’s data to a constant

set_rand_uniform()¶: Sets this vector’s elements to values from random uniform

set_randn()¶: Sets this vector’s elements to random [0,1)

set_zero()¶: Sets this vector’s data to zero

sum() → float¶: Sum this vector’s elements

swap(vec:DoubleVector)¶: Swaps the contents of this and vec. Shallow swap.

write(os:ostream, binary:bool)¶: I/O

class kaldi.cudamatrix.CuFloatPtrArray¶

CLIF wrapper for ::kaldi::CuArray<float *>

copy_from_array(src:CuFloatPtrArray)¶: This function resizes if needed.

destroy()¶: Deallocate the memory and set dim_ and data_ to zero.

dim() → int¶

Returns:	This vector dimension

from_other(src:CuFloatPtrArray) → CuFloatPtrArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray<float *>> ::kaldi::CuArray<float *>::CuArray(::kaldi::CuArray<float *>)

from_size(dim:int, resize_type:MatrixResizeType=default) → CuFloatPtrArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray<float *>> ::kaldi::CuArray<float *>::CuArray(int, ::kaldi::MatrixResizeType)

max() → float¶

Returns:	Maximum value

min() → float¶

Returns:	Minimum value

resize(dim:int, resize_type:MatrixResizeType=default)¶: Allocate the memory.

set_zero()¶: Sets the memory for the object to zero.

class kaldi.cudamatrix.CuIntPairArray¶

CLIF wrapper for ::kaldi::CuArray< ::Int32Pair>

add(value:IntPair)¶: Add a constant value

copy_from_array(src:CuIntPairArray)¶: This function resizes if needed.

copy_from_vec(src:list<IntPair>)¶: This function resizes if needed.

copy_to_vec() → list<IntPair>¶: This function resizes dst if needed.

destroy()¶: Deallocate the memory and set dim_ and data_ to zero.

dim() → int¶

Returns:	This vector dimension

from_other(src:CuIntPairArray) → CuIntPairArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray< ::Int32Pair>> ::kaldi::CuArray<Int32Pair>::CuArray(::kaldi::CuArray< ::Int32Pair>)

from_size(dim:int, resize_type:MatrixResizeType=default) → CuIntPairArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray< ::Int32Pair>> ::kaldi::CuArray<Int32Pair>::CuArray(int, ::kaldi::MatrixResizeType)

from_vector(src:list<IntPair>) → CuIntPairArray¶: Calls C++ function std::unique_ptr<::kaldi::CuArray< ::Int32Pair>> ::kaldi::CuArray<Int32Pair>::CuArray(::std::vector< ::Int32Pair>)

resize(dim:int, resize_type:MatrixResizeType=default)¶: Allocate the memory.

set(value:IntPair)¶: Set to a constant value.

set_zero()¶: Sets the memory for the object to zero.

class kaldi.cudamatrix.CuMatrix¶

CLIF wrapper for ::kaldi::CuMatrix<float>

add(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Add(float)

add_cols(src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddCols(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

add_diag_vec_mat(alpha:float, v:CuVectorBase, M:CuMatrixBase, transM:MatrixTransposeType, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddDiagVecMat(float, ::kaldi::CuVectorBase<float>, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, float)

add_elements(alpha:float, input:list<MatrixElement>)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddElements(float, ::std::vector< ::MatrixElement<float> >)

add_mat(alpha:float, A:CuMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMat(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType)

add_mat_blocks(alpha:float, A:CuMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMatBlocks(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType)

add_mat_diag_vec(alpha:float, M:CuMatrixBase, transM:MatrixTransposeType, v:CuVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMatDiagVec(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, ::kaldi::CuVectorBase<float>, float)

add_mat_mat(alpha:float, A:CuMatrixBase, transA:MatrixTransposeType, B:CuMatrixBase, transB:MatrixTransposeType, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMatMat(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, float)

add_mat_mat_elements(alpha:float, A:CuMatrixBase, B:CuMatrixBase, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMatMatElements(float, ::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, float)

add_row_ranges(src:CuMatrixBase, indexes:CuIntPairArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddRowRanges(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray< ::Int32Pair>)

add_rows(alpha:float, src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddRows(float, ::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

add_to_diag(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddToDiag(float)

add_to_rows(alpha:float, dst:CuFloatPtrArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddToRows(float, ::kaldi::CuArray<float *>)

add_vec_to_cols(alpha:float, col:CuVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddVecToCols(float, ::kaldi::CuVectorBase<float>, float)

add_vec_to_rows(alpha:float, row:CuVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddVecToRows(float, ::kaldi::CuVectorBase<float>, float)

add_vec_vec(alpha:float, x:CuVectorBase, y:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddVecVec(float, ::kaldi::CuVectorBase<float>, ::kaldi::CuVectorBase<float>)

apply_ceiling(ceiling_val:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyCeiling(float)

apply_exp()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyExp()

apply_exp_limited(lower_limit:float, upper_limit:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyExpLimited(float, float)

apply_floor(floor_val:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyFloor(float)

apply_heaviside()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyHeaviside()

apply_log()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyLog()

apply_log_soft_max_per_row(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyLogSoftMaxPerRow(::kaldi::CuMatrixBase<float>)

apply_pow(power:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyPow(float)

apply_pow_abs(power:float, include_sign:bool=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyPowAbs(float, bool)

apply_soft_max_per_row(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplySoftMaxPerRow(::kaldi::CuMatrixBase<float>)

approx_equal(other:CuMatrixBase, tol:float=default) → bool¶: Calls C++ function bool ::kaldi::CuMatrixBase<float>::ApproxEqual(::kaldi::CuMatrixBase<float>, float)

cholesky(inv_cholesky:CuMatrixBase=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Cholesky(::kaldi::CuMatrixBase<float> *)

comp_objf_and_deriv(elements:list<MatrixElement>, A:CuMatrix) -> (tot_objf:float, tot_weight:float)¶: Calls C++ function void ::kaldi::CuMatrix<float>::CompObjfAndDeriv(::std::vector< ::MatrixElement<float> >, ::kaldi::CuMatrix<float>, float*, float*)

copy_col_from_cu_vec(v:CuVectorBase, col:int)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyColFromVec(::kaldi::CuVectorBase<float>, int)

copy_cols(src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyCols(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

copy_cols_from_cu_vec(v:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyColsFromVec(::kaldi::CuVectorBase<float>)

copy_from_cu_mat(src:CuMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyFromMat(::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType)

copy_from_mat(src:MatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyFromMat(::kaldi::MatrixBase<float>, ::kaldi::MatrixTransposeType)

copy_lower_to_upper()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyLowerToUpper()

copy_rows(src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyRows(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

copy_rows_from_cu_vec(v:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyRowsFromVec(::kaldi::CuVectorBase<float>)

copy_rows_from_vec(v:VectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyRowsFromVec(::kaldi::VectorBase<float>)

copy_to_mat(dst:MatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyToMat(::kaldi::MatrixBase<float> *, ::kaldi::MatrixTransposeType)

copy_to_rows(dst:CuFloatPtrArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyToRows(::kaldi::CuArray<float *>)

copy_upper_to_lower()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyUpperToLower()

diff_group_pnorm(in_value:CuMatrixBase, out_value:CuMatrixBase, out_deriv:CuMatrixBase, power:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffGroupPnorm(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, float)

diff_log_softmax_per_row(out_value:CuMatrixBase, out_deriv:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffLogSoftmaxPerRow(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

diff_parametric_relu(value:CuMatrixBase, diff:CuMatrixBase, alpha:CuVectorBase, beta:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffParametricRelu(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, ::kaldi::CuVectorBase<float>, ::kaldi::CuVectorBase<float>)

diff_sigmoid(value:CuMatrixBase, diff:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffSigmoid(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

diff_softmax_per_row(value:CuMatrixBase, diff:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffSoftmaxPerRow(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

diff_tanh(value:CuMatrixBase, diff:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffTanh(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

diff_xent(tgt:CuArray) → CuVector¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffXent(::kaldi::CuArray<int>, ::kaldi::CuVector<float>*)

dim() → MatrixDim¶: Calls C++ function ::MatrixDim_ ::kaldi::CuMatrixBase<float>::Dim()

div_elements(A:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DivElements(::kaldi::CuMatrixBase<float>)

div_rows_vec(div:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DivRowsVec(::kaldi::CuVectorBase<float>)

find_row_max_id() → CuArray¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::FindRowMaxId(::kaldi::CuArray<int>*)

frobenius_norm() → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::FrobeniusNorm()

from_matrix(other:MatrixBase, trans:MatrixTransposeType=default) → CuMatrix¶: Calls C++ function std::unique_ptr<::kaldi::CuMatrix<float>> ::kaldi::CuMatrix<float>::CuMatrix(::kaldi::MatrixBase<float>, ::kaldi::MatrixTransposeType)

from_other(other:CuMatrixBase, trans:MatrixTransposeType=default) → CuMatrix¶: Calls C++ function std::unique_ptr<::kaldi::CuMatrix<float>> ::kaldi::CuMatrix<float>::CuMatrix(::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType)

from_size(rows:int, cols:int, resize_type:MatrixResizeType=default, stride_type:MatrixStrideType=default) → CuMatrix¶: Calls C++ function std::unique_ptr<::kaldi::CuMatrix<float>> ::kaldi::CuMatrix<float>::CuMatrix(int, int, ::kaldi::MatrixResizeType, ::kaldi::MatrixStrideType)

group_max(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::GroupMax(::kaldi::CuMatrixBase<float>)

group_max_deriv(input:CuMatrixBase, output:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::GroupMaxDeriv(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

group_pnorm(src:CuMatrixBase, pow:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::GroupPnorm(::kaldi::CuMatrixBase<float>, float)

heaviside(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Heaviside(::kaldi::CuMatrixBase<float>)

invert_elements()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::InvertElements()

is_unit(tol:float=default) → bool¶: Calls C++ function bool ::kaldi::CuMatrixBase<float>::IsUnit(float)

max() → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::Max()

max_elementwise(A:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Max(::kaldi::CuMatrixBase<float>)

min() → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::Min()

min_elementwise(A:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Min(::kaldi::CuMatrixBase<float>)

mul_cols_vec(scale:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulColsVec(::kaldi::CuVectorBase<float>)

mul_elements(A:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulElements(::kaldi::CuMatrixBase<float>)

mul_rows(src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulRows(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

mul_rows_group_mat(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulRowsGroupMat(::kaldi::CuMatrixBase<float>)

mul_rows_vec(scale:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulRowsVec(::kaldi::CuVectorBase<float>)

num_cols() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<float>::NumCols()

num_rows() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<float>::NumRows()

parametric_relu(src:CuMatrixBase, alpha:CuVectorBase, beta:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ParametricRelu(::kaldi::CuMatrixBase<float>, ::kaldi::CuVectorBase<float>, ::kaldi::CuVectorBase<float>)

read(is:istream, binary:bool)¶: Calls C++ function void ::kaldi::CuMatrix<float>::Read(::std::basic_istream<char, ::std::char_traits<char> >, bool)

resize(rows:int, cols:int, resize_type:MatrixResizeType=default, stride_type:MatrixStrideType=default)¶: Calls C++ function void ::kaldi::CuMatrix<float>::Resize(int, int, ::kaldi::MatrixResizeType, ::kaldi::MatrixStrideType)

scale(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Scale(float)

set(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Set(float)

set_mat_mat_div_mat(A:CuMatrixBase, B:CuMatrixBase, C:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetMatMatDivMat(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

set_rand_uniform()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetRandUniform()

set_randn()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetRandn()

set_zero()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetZero()

set_zero_above_diag()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetZeroAboveDiag()

sigmoid(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Sigmoid(::kaldi::CuMatrixBase<float>)

size_in_bytes() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<float>::SizeInBytes()

soft_hinge(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SoftHinge(::kaldi::CuMatrixBase<float>)

stride() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<float>::Stride()

sum() → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::Sum()

sum_column_ranges(src:CuMatrixBase, indexes:CuIntPairArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SumColumnRanges(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray< ::Int32Pair>)

swap(mat:CuMatrix)¶: Calls C++ function void ::kaldi::CuMatrix<float>::Swap(::kaldi::CuMatrix<float> *)

swap_with_matrix(mat:Matrix)¶: Calls C++ function void ::kaldi::CuMatrix<float>::Swap(::kaldi::Matrix<float> *)

sym_add_mat2(alpha:float, M:CuMatrixBase, transA:MatrixTransposeType, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SymAddMat2(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, float)

sym_invert_pos_def()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SymInvertPosDef()

tanh(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Tanh(::kaldi::CuMatrixBase<float>)

trace(check_square:bool=default) → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::Trace(bool)

transpose()¶: Calls C++ function void ::kaldi::CuMatrix<float>::Transpose()

write(os:ostream, binary:bool)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Write(::std::basic_ostream<char, ::std::char_traits<char> >, bool)

class kaldi.cudamatrix.CuSparseMatrix¶

This class defines a Sparse CUDA Matrix.

copy_elements_to_vec(vec:CuVectorBase)¶

Copies the nonzero elements of the matrix to the vector other.

Parameters:	vec (CuVectorBase) – A CUDA vector to store the nonzero elements of self
Raises:	Error if the size of other is different than the number of nonzero elements of self

copy_from_smat_(smat:CuSparseMatrix, trans:MatrixTransposeType=default)¶

Copy data from another sparse CUDA matrix. Will resize if necessary.

Parameters:	smat (CuSparseMatrix) – The sparse CUDA matrix to copy from. trans (MatrixTransposeType) – transpose or not

copy_to_mat(dest:CuMatrixBase, trans:MatrixTransposeType=default)¶

Copies elements to a CUDA matrix.

Parameters:	dest (CuMatrixBase) – The matrix to copy to. trans (MatrixTransposeType) – transpose or not.

copy_to_smat(smat:SparseMatrix)¶

Copy data to a sparse matrix.

Parameters:	smat (SparseMatrix) – The sparse matrix to copy to.

frobenius_norm() → float¶: Computes and returns the Frobenius norm.

from_cpu(smat:SparseMatrix) → CuSparseMatrix¶

Initialization by copying the elements of a sparse matrix.

Parameters:	smat (SparseMatrix) – A sparse matrix.

from_indexes(indexes:CuArray, dim:int, trans:MatrixTransposeType=default) → CuSparseMatrix¶

Initialization from an array of indexes.

If trans == NO_TRANS, construct a sparse matrix with num-rows == indexes.Dim() and num-cols = ‘dim’. ‘indexes’ is expected to contain elements in the range [0, dim - 1]. Each row ‘i’ of self after calling the constructor will contain a single element at column-index indexes[i] with value 1.0.

If trans == TRANS, the result will be the transpose of the sparse matrix described above.

Parameters:	indexes (CuArray) – The CUDA array of indexes. dim (int) – The column (or row) dimension. trans (MatrixTransposeType) – Transpose or not.

from_indexes_and_weights(indexes:CuArray, weights:CuVectorBase, dim:int, trans:MatrixTransposeType=default) → CuSparseMatrix¶

Initialization from an array of indexes and an array of weights.

Requires indexes.Dim() == weights.Dim(). If trans == NO_TRANS, construct a sparse matrix with num-rows == indexes.Dim() and num-cols = ‘dim’. ‘indexes’ is expected to contain elements in the range [0, dim - 1]. Each row ‘i’ of *this after calling the constructor will contain a single element at column-index indexes[i] with value weights[i]. If trans == TRANS, the result will be the transpose of the sparse matrix described above.

Parameters:	indexes (CuArray) – The CUDA array of indexes. weights (CuVectorBase) – The CUDA vector of weights. dim (int) – The column (or row) dimension. trans (MatrixTransposeType) – Transpose or not.

from_other(smat:CuSparseMatrix, trans:MatrixTransposeType=default) → CuSparseMatrix¶

Initialization by copying the elements of another sparse CUDA matrix.

Parameters:	smat (CuSparseMatrix) – A sparse CUDA matrix trans (MatrixTransposeType) – transpose or not

num_cols¶: Number of columns (zero for empty matrix).

num_elements() → int¶: Returns the number of nonzero elements.

num_rows¶: Number of rows (zero for empty matrix).

read_(is:istream, binary:bool)¶

Reads from C++ stream.

Parameters:	is (istream) – The input straem. binary (bool) – If the stream is binary.

select_rows_(row_indexes:CuArray, smat_other:CuSparseMatrix)¶

Sets self to a subset of the rows of another sparse CUDA matrix.

Parameters:	row_indexes (CuArray) – The rows to select. smat_other (CuSparseMatrix) – The other sparse CUDA matrix.

set_randn_(zero_prob:float)¶: Sets up to a pseudo-randomly initialized matrix, with each element zero with probability zero_prob and else normally distributed.

sum() → float¶: Returns the sum of the elememts.

swap_(other:CuSparseMatrix)¶

Swaps the contents with another sparse CUDA matrix.

Shallow swap.

swap_with_cpu_(other:SparseMatrix)¶

Swaps the contents with a sparse matrix.

Shallow swap.

write(os:ostream, binary:bool)¶

Writes to C++ stream.

Parameters:	os (ostream) – The output straem. binary (bool) – If the stream is binary.

class kaldi.cudamatrix.CuSubMatrix¶

CLIF wrapper for ::kaldi::CuSubMatrix<float>

add(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Add(float)

add_cols(src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddCols(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

add_diag_vec_mat(alpha:float, v:CuVectorBase, M:CuMatrixBase, transM:MatrixTransposeType, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddDiagVecMat(float, ::kaldi::CuVectorBase<float>, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, float)

add_elements(alpha:float, input:list<MatrixElement>)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddElements(float, ::std::vector< ::MatrixElement<float> >)

add_mat(alpha:float, A:CuMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMat(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType)

add_mat_blocks(alpha:float, A:CuMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMatBlocks(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType)

add_mat_diag_vec(alpha:float, M:CuMatrixBase, transM:MatrixTransposeType, v:CuVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMatDiagVec(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, ::kaldi::CuVectorBase<float>, float)

add_mat_mat(alpha:float, A:CuMatrixBase, transA:MatrixTransposeType, B:CuMatrixBase, transB:MatrixTransposeType, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMatMat(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, float)

add_mat_mat_elements(alpha:float, A:CuMatrixBase, B:CuMatrixBase, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddMatMatElements(float, ::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, float)

add_row_ranges(src:CuMatrixBase, indexes:CuIntPairArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddRowRanges(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray< ::Int32Pair>)

add_rows(alpha:float, src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddRows(float, ::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

add_to_diag(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddToDiag(float)

add_to_rows(alpha:float, dst:CuFloatPtrArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddToRows(float, ::kaldi::CuArray<float *>)

add_vec_to_cols(alpha:float, col:CuVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddVecToCols(float, ::kaldi::CuVectorBase<float>, float)

add_vec_to_rows(alpha:float, row:CuVectorBase, beta:float=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddVecToRows(float, ::kaldi::CuVectorBase<float>, float)

add_vec_vec(alpha:float, x:CuVectorBase, y:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::AddVecVec(float, ::kaldi::CuVectorBase<float>, ::kaldi::CuVectorBase<float>)

apply_ceiling(ceiling_val:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyCeiling(float)

apply_exp()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyExp()

apply_exp_limited(lower_limit:float, upper_limit:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyExpLimited(float, float)

apply_floor(floor_val:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyFloor(float)

apply_heaviside()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyHeaviside()

apply_log()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyLog()

apply_log_soft_max_per_row(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyLogSoftMaxPerRow(::kaldi::CuMatrixBase<float>)

apply_pow(power:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyPow(float)

apply_pow_abs(power:float, include_sign:bool=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplyPowAbs(float, bool)

apply_soft_max_per_row(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ApplySoftMaxPerRow(::kaldi::CuMatrixBase<float>)

approx_equal(other:CuMatrixBase, tol:float=default) → bool¶: Calls C++ function bool ::kaldi::CuMatrixBase<float>::ApproxEqual(::kaldi::CuMatrixBase<float>, float)

cholesky(inv_cholesky:CuMatrixBase=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Cholesky(::kaldi::CuMatrixBase<float> *)

copy_col_from_cu_vec(v:CuVectorBase, col:int)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyColFromVec(::kaldi::CuVectorBase<float>, int)

copy_cols(src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyCols(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

copy_cols_from_cu_vec(v:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyColsFromVec(::kaldi::CuVectorBase<float>)

copy_from_cu_mat(src:CuMatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyFromMat(::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType)

copy_from_mat(src:MatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyFromMat(::kaldi::MatrixBase<float>, ::kaldi::MatrixTransposeType)

copy_lower_to_upper()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyLowerToUpper()

copy_rows(src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyRows(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

copy_rows_from_cu_vec(v:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyRowsFromVec(::kaldi::CuVectorBase<float>)

copy_rows_from_vec(v:VectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyRowsFromVec(::kaldi::VectorBase<float>)

copy_to_mat(dst:MatrixBase, trans:MatrixTransposeType=default)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyToMat(::kaldi::MatrixBase<float> *, ::kaldi::MatrixTransposeType)

copy_to_rows(dst:CuFloatPtrArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyToRows(::kaldi::CuArray<float *>)

copy_upper_to_lower()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::CopyUpperToLower()

diff_group_pnorm(in_value:CuMatrixBase, out_value:CuMatrixBase, out_deriv:CuMatrixBase, power:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffGroupPnorm(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, float)

diff_log_softmax_per_row(out_value:CuMatrixBase, out_deriv:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffLogSoftmaxPerRow(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

diff_parametric_relu(value:CuMatrixBase, diff:CuMatrixBase, alpha:CuVectorBase, beta:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffParametricRelu(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, ::kaldi::CuVectorBase<float>, ::kaldi::CuVectorBase<float>)

diff_sigmoid(value:CuMatrixBase, diff:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffSigmoid(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

diff_softmax_per_row(value:CuMatrixBase, diff:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffSoftmaxPerRow(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

diff_tanh(value:CuMatrixBase, diff:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffTanh(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

diff_xent(tgt:CuArray) → CuVector¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DiffXent(::kaldi::CuArray<int>, ::kaldi::CuVector<float>*)

dim() → MatrixDim¶: Calls C++ function ::MatrixDim_ ::kaldi::CuMatrixBase<float>::Dim()

div_elements(A:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DivElements(::kaldi::CuMatrixBase<float>)

div_rows_vec(div:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::DivRowsVec(::kaldi::CuVectorBase<float>)

find_row_max_id() → CuArray¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::FindRowMaxId(::kaldi::CuArray<int>*)

frobenius_norm() → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::FrobeniusNorm()

from_other(other:CuSubMatrix) → CuSubMatrix¶: Calls C++ function std::unique_ptr<::kaldi::CuSubMatrix<float>> ::kaldi::CuSubMatrix<float>::CuSubMatrix(::kaldi::CuSubMatrix<float>)

group_max(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::GroupMax(::kaldi::CuMatrixBase<float>)

group_max_deriv(input:CuMatrixBase, output:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::GroupMaxDeriv(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

group_pnorm(src:CuMatrixBase, pow:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::GroupPnorm(::kaldi::CuMatrixBase<float>, float)

heaviside(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Heaviside(::kaldi::CuMatrixBase<float>)

invert_elements()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::InvertElements()

is_unit(tol:float=default) → bool¶: Calls C++ function bool ::kaldi::CuMatrixBase<float>::IsUnit(float)

max() → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::Max()

max_elementwise(A:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Max(::kaldi::CuMatrixBase<float>)

min() → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::Min()

min_elementwise(A:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Min(::kaldi::CuMatrixBase<float>)

mul_cols_vec(scale:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulColsVec(::kaldi::CuVectorBase<float>)

mul_elements(A:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulElements(::kaldi::CuMatrixBase<float>)

mul_rows(src:CuMatrixBase, indexes:CuArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulRows(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray<int>)

mul_rows_group_mat(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulRowsGroupMat(::kaldi::CuMatrixBase<float>)

mul_rows_vec(scale:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::MulRowsVec(::kaldi::CuVectorBase<float>)

num_cols() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<float>::NumCols()

num_rows() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<float>::NumRows()

parametric_relu(src:CuMatrixBase, alpha:CuVectorBase, beta:CuVectorBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::ParametricRelu(::kaldi::CuMatrixBase<float>, ::kaldi::CuVectorBase<float>, ::kaldi::CuVectorBase<float>)

scale(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Scale(float)

set(value:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Set(float)

set_mat_mat_div_mat(A:CuMatrixBase, B:CuMatrixBase, C:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetMatMatDivMat(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

set_rand_uniform()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetRandUniform()

set_randn()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetRandn()

set_zero()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetZero()

set_zero_above_diag()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SetZeroAboveDiag()

sigmoid(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Sigmoid(::kaldi::CuMatrixBase<float>)

size_in_bytes() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<float>::SizeInBytes()

soft_hinge(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SoftHinge(::kaldi::CuMatrixBase<float>)

stride() → int¶: Calls C++ function int ::kaldi::CuMatrixBase<float>::Stride()

sum() → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::Sum()

sum_column_ranges(src:CuMatrixBase, indexes:CuIntPairArray)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SumColumnRanges(::kaldi::CuMatrixBase<float>, ::kaldi::CuArray< ::Int32Pair>)

sym_add_mat2(alpha:float, M:CuMatrixBase, transA:MatrixTransposeType, beta:float)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SymAddMat2(float, ::kaldi::CuMatrixBase<float>, ::kaldi::MatrixTransposeType, float)

sym_invert_pos_def()¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::SymInvertPosDef()

tanh(src:CuMatrixBase)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Tanh(::kaldi::CuMatrixBase<float>)

trace(check_square:bool=default) → float¶: Calls C++ function float ::kaldi::CuMatrixBase<float>::Trace(bool)

write(os:ostream, binary:bool)¶: Calls C++ function void ::kaldi::CuMatrixBase<float>::Write(::std::basic_ostream<char, ::std::char_traits<char> >, bool)

class kaldi.cudamatrix.CuSubVector¶

CLIF wrapper for ::kaldi::CuSubVector<float>

add(value:float)¶: Adds a constant to this vector

add_vec(alpha:float, vec:CuVectorBase, beta:float=default)¶: Computes this <– beta * this + alpha * vec

add_vec_vec(alpha:float, v:CuVectorBase, r:CuVectorBase, beta:float)¶: Computes this <– alpha * v + beta * r

apply_ceiling(ceiling_val:float) → int¶

Applies ceiling to this vector’s elements

Returns:	The number of elements set to ceiling value.

apply_ceiling_no_count(ceiling_val:float)¶: Applies ceiling to this vector’s elements

apply_exp()¶: Applies exp to this vector’s elements

apply_floor(floor_val:float) → int¶

Applies floor to this vector’s elements

Returns:	The number of elements set to floor value.

apply_floor_no_count(floor_val:float)¶: Applies floor to this vector’s elements

apply_log()¶: Applies log to this vector’s elements

apply_pow(power:float)¶: Applies pow to this vector’s elements

apply_softmax()¶: Applies a softmax to this vector’s elements

approx_equal(other:CuVectorBase, tol:float=default) → bool¶: Returns true if ||this - other|| <= tol

copy_from_cu_vec(src:CuVectorBase)¶: Copies data from other CuVector

copy_from_vec(src:VectorBase)¶: Copies data from other :class:kaldi.matrix.VectorBase

copy_rows_from_mat(M:MatrixBase)¶: Copies data from :class:kaldi.matrix.MatrixBase

copy_to_vec(dst:VectorBase)¶: Copies data to :class:kaldi.matrix.VectorBase

dim() → int¶: Returns this vector dimension

from_other(other:CuSubVector) → CuSubVector¶: Calls C++ function std::unique_ptr<::kaldi::CuSubVector<float>> ::kaldi::CuSubVector<float>::CuSubVector(::kaldi::CuSubVector<float>)

invert_elements()¶: Inverts this vector’s elements

max() → float¶

Returns:	Maximum value of this vector

min() → float¶

Returns:	Minimum value of this vector

mul_elements(v:CuVectorBase)¶: Multipy element-by-element by another vector.

norm(p:float) → float¶

Returns:	p-norm of this vector

replace_value(orig:float, changed:float)¶: Set each element to y = (x == orig ? changed : x).

scale(value:float)¶: Multiplies this vector’s data by a constant

set(value:float)¶: Sets this vector’s data to a constant

set_rand_uniform()¶: Sets this vector’s elements to values from random uniform

set_randn()¶: Sets this vector’s elements to random [0,1)

set_zero()¶: Sets this vector’s data to zero

sum() → float¶: Sum this vector’s elements

class kaldi.cudamatrix.CuTimer¶

Convenience wrapper for Timer.

This class sets the time only if the verbose level is >= 1. This helps avoid an unnecessary system call if the verbose level is 0 and you won’t be accumulating the timing stats.

elapsed() → float¶: Returns elapsed time since timer was reset.

new(set_timer:bool) → Timer¶: Creates and optionally resets the timer to current time.

reset()¶: Resets the timer to current time.

class kaldi.cudamatrix.CuVector¶

CLIF wrapper for ::kaldi::CuVector<float>

add(value:float)¶: Adds a constant to this vector

add_vec(alpha:float, vec:CuVectorBase, beta:float=default)¶: Computes this <– beta * this + alpha * vec

add_vec_vec(alpha:float, v:CuVectorBase, r:CuVectorBase, beta:float)¶: Computes this <– alpha * v + beta * r

apply_ceiling(ceiling_val:float) → int¶

Applies ceiling to this vector’s elements

Returns:	The number of elements set to ceiling value.

apply_ceiling_no_count(ceiling_val:float)¶: Applies ceiling to this vector’s elements

apply_exp()¶: Applies exp to this vector’s elements

apply_floor(floor_val:float) → int¶

Applies floor to this vector’s elements

Returns:	The number of elements set to floor value.

apply_floor_no_count(floor_val:float)¶: Applies floor to this vector’s elements

apply_log()¶: Applies log to this vector’s elements

apply_pow(power:float)¶: Applies pow to this vector’s elements

apply_softmax()¶: Applies a softmax to this vector’s elements

approx_equal(other:CuVectorBase, tol:float=default) → bool¶: Returns true if ||this - other|| <= tol

copy_from_cu_vec(src:CuVectorBase)¶: Copies data from other CuVector

copy_from_vec(src:VectorBase)¶: Copies data from other :class:kaldi.matrix.VectorBase

copy_rows_from_mat(M:MatrixBase)¶: Copies data from :class:kaldi.matrix.MatrixBase

copy_to_vec(dst:VectorBase)¶: Copies data to :class:kaldi.matrix.VectorBase

dim() → int¶: Returns this vector dimension

from_base(v:CuVectorBase) → CuVector¶: Calls C++ function std::unique_ptr<::kaldi::CuVector<float>> ::kaldi::CuVector<float>::CuVector(::kaldi::CuVectorBase<float>)

from_size(dim:int, resize_type:MatrixResizeType=default) → CuVector¶: Calls C++ function std::unique_ptr<::kaldi::CuVector<float>> ::kaldi::CuVector<float>::CuVector(int, ::kaldi::MatrixResizeType)

from_vector(v:VectorBase) → CuVector¶: Calls C++ function std::unique_ptr<::kaldi::CuVector<float>> ::kaldi::CuVector<float>::CuVector(::kaldi::VectorBase<float>)

invert_elements()¶: Inverts this vector’s elements

max() → float¶

Returns:	Maximum value of this vector

min() → float¶

Returns:	Minimum value of this vector

mul_elements(v:CuVectorBase)¶: Multipy element-by-element by another vector.

norm(p:float) → float¶

Returns:	p-norm of this vector

read(is:istream, binary:bool)¶: I/O

replace_value(orig:float, changed:float)¶: Set each element to y = (x == orig ? changed : x).

resize(dim:int, resize_type:MatrixResizeType=default)¶: Resize this vector to dim

scale(value:float)¶: Multiplies this vector’s data by a constant

set(value:float)¶: Sets this vector’s data to a constant

set_rand_uniform()¶: Sets this vector’s elements to values from random uniform

set_randn()¶: Sets this vector’s elements to random [0,1)

set_zero()¶: Sets this vector’s data to zero

sum() → float¶: Sum this vector’s elements

swap(vec:Vector)¶: Swaps the contents of this and vec. Shallow swap.

write(os:ostream, binary:bool)¶: I/O

class kaldi.cudamatrix.DoubleMatrixElement¶

CLIF wrapper for ::MatrixElement<double>

column¶: C++ ::int32_cuda MatrixElement<double>.column

row¶: C++ ::int32_cuda MatrixElement<double>.row

weight¶: C++ double MatrixElement<double>.weight

class kaldi.cudamatrix.IntPair¶

CLIF wrapper for ::Int32Pair

first¶: C++ ::int32_cuda Int32Pair.first

second¶: C++ ::int32_cuda Int32Pair.second

class kaldi.cudamatrix.MatrixDim¶

CLIF wrapper for ::MatrixDim_

cols¶: C++ ::int32_cuda MatrixDim_.cols

rows¶: C++ ::int32_cuda MatrixDim_.rows

stride¶: C++ ::int32_cuda MatrixDim_.stride

class kaldi.cudamatrix.MatrixElement¶

CLIF wrapper for ::MatrixElement<float>

column¶: C++ ::int32_cuda MatrixElement<float>.column

row¶: C++ ::int32_cuda MatrixElement<float>.row

weight¶: C++ float MatrixElement<float>.weight

kaldi.cudamatrix.approx_equal_cu_double_matrix(A:CuDoubleMatrixBase, B:CuDoubleMatrixBase, tol:float=default) → bool¶: Calls C++ function bool ::kaldi::ApproxEqual(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, double)

kaldi.cudamatrix.approx_equal_cu_double_vector(a:CuDoubleVectorBase, b:CuDoubleVectorBase, tol:float=default) → bool¶: Calls C++ function bool ::kaldi::ApproxEqual(::kaldi::CuVectorBase<double>, ::kaldi::CuVectorBase<double>, double)

kaldi.cudamatrix.approx_equal_cu_matrix(A:CuMatrixBase, B:CuMatrixBase, tol:float=default) → bool¶: Calls C++ function bool ::kaldi::ApproxEqual(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, float)

kaldi.cudamatrix.approx_equal_cu_vector(a:CuVectorBase, b:CuVectorBase, tol:float=default) → bool¶: Calls C++ function bool ::kaldi::ApproxEqual(::kaldi::CuVectorBase<float>, ::kaldi::CuVectorBase<float>, float)

kaldi.cudamatrix.assert_equal_cu_double_matrix(A:CuDoubleMatrixBase, B:CuDoubleMatrixBase, tol:float=default)¶: Calls C++ function void ::kaldi::AssertEqual(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>, float)

kaldi.cudamatrix.assert_equal_cu_double_vector(a:CuDoubleVectorBase, b:CuDoubleVectorBase, tol:float=default)¶: Calls C++ function void ::kaldi::AssertEqual(::kaldi::CuVectorBase<double>, ::kaldi::CuVectorBase<double>, double)

kaldi.cudamatrix.assert_equal_cu_matrix(A:CuMatrixBase, B:CuMatrixBase, tol:float=default)¶: Calls C++ function void ::kaldi::AssertEqual(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>, float)

kaldi.cudamatrix.assert_equal_cu_vector(a:CuVectorBase, b:CuVectorBase, tol:float=default)¶: Calls C++ function void ::kaldi::AssertEqual(::kaldi::CuVectorBase<float>, ::kaldi::CuVectorBase<float>, float)

kaldi.cudamatrix.cuda_available()[source]¶: Check if cuda is available.

kaldi.cudamatrix.same_dim_and_stride_cu_double_matrix(M:CuDoubleMatrixBase, N:CuDoubleMatrixBase) → bool¶: Calls C++ function bool ::kaldi::SameDimAndStride(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

kaldi.cudamatrix.same_dim_and_stride_cu_matrix(M:CuMatrixBase, N:CuMatrixBase) → bool¶: Calls C++ function bool ::kaldi::SameDimAndStride(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

kaldi.cudamatrix.same_dim_cu_double_matrix(M:CuDoubleMatrixBase, N:CuDoubleMatrixBase) → bool¶: Calls C++ function bool ::kaldi::SameDim(::kaldi::CuMatrixBase<double>, ::kaldi::CuMatrixBase<double>)

kaldi.cudamatrix.same_dim_cu_matrix(M:CuMatrixBase, N:CuMatrixBase) → bool¶: Calls C++ function bool ::kaldi::SameDim(::kaldi::CuMatrixBase<float>, ::kaldi::CuMatrixBase<float>)

kaldi.cudamatrix.synchronize_gpu()¶: Synchronizes threads.

kaldi.cudamatrix.trace_mat_smat(A:CuMatrixBase, B:CuSparseMatrix, trans:MatrixTransposeType=default) → float¶

If trans==kNoTrans compute Trace(A*B). If trans==kTrans compute Trace(A^T*B)

Parameters:	A (CuMatrix) – A CUDA matrix B (CuSparseMatrix) – A sparse CUDA matrix trans (MatrixTransposeType) – Determines if we use A as is or its transpose
Raises:	Error in case of dimensionality mismatch