public CudaProvider(string cudaKernelPath, bool stochastic = true)
{
_stochastic = stochastic;
_cuda = new CudaContext();
_kernel = new KernelModule(_cuda, cudaKernelPath);
_blas = new CudaBlas(AtomicsMode.Allowed);
_pointwiseMultiply = _kernel.LoadFunction("PointwiseMultiply");
_addInPlace = _kernel.LoadFunction("AddInPlace");
_subtractInPlace = _kernel.LoadFunction("SubtractInPlace");
_addToEachRow = _kernel.LoadFunction("AddToEachRow");
_addToEachColumn = _kernel.LoadFunction("AddToEachColumn");
_tanh = _kernel.LoadFunction("TanH");
_tanhDerivative = _kernel.LoadFunction("TanHDerivative");
_sigmoid = _kernel.LoadFunction("Sigmoid");
_sigmoidDerivative = _kernel.LoadFunction("SigmoidDerivative");
_sumRows = _kernel.LoadFunction("SumRows");
_relu = _kernel.LoadFunction("RELU");
_reluDerivative = _kernel.LoadFunction("RELUDerivative");
_memClear = _kernel.LoadFunction("MemClear");
_sparseLoad = _kernel.LoadFunction("SparseLoad");
_sumColumns = _kernel.LoadFunction("SumColumns");
_pointwiseDivide = _kernel.LoadFunction("PointwiseDivide");
_sqrt = _kernel.LoadFunction("Sqrt");
_memCopy = _kernel.LoadFunction("MemCopy");
_findMinAndMax = _kernel.LoadFunction("FindMinAndMax");
_findSum = _kernel.LoadFunction("FindSum");
_findStdDev = _kernel.LoadFunction("FindStdDev");
_constrain = _kernel.LoadFunction("Constrain");
_pow = _kernel.LoadFunction("Pow");
_diagonal = _kernel.LoadFunction("Diagonal");
_l1Regularisation = _kernel.LoadFunction("L1Regularisation");
_leakyRelu = _kernel.LoadFunction("LeakyRELU");
_leakyReluDerivative = _kernel.LoadFunction("LeakyRELUDerivative");
_pointwiseDivideRows = _kernel.LoadFunction("PointwiseDivideRows");
_pointwiseDivideColumns = _kernel.LoadFunction("PointwiseDivideColumns");
_splitRows = _kernel.LoadFunction("SplitRows");
_splitColumns = _kernel.LoadFunction("SplitColumns");
_concatRows = _kernel.LoadFunction("ConcatRows");
_concatColumns = _kernel.LoadFunction("ConcatColumns");
_euclideanDistance = _kernel.LoadFunction("EuclideanDistance");
_manhattanDistance = _kernel.LoadFunction("ManhattanDistance");
_abs = _kernel.LoadFunction("Abs");
_normalise = _kernel.LoadFunction("Normalise");
_softmaxVector = _kernel.LoadFunction("SoftmaxVector");
_multiEuclidean = _kernel.LoadFunction("MultiEuclideanDistance");
_multiManhattan = _kernel.LoadFunction("MultiManhattanDistance");
_log = _kernel.LoadFunction("Log");
_sigmoidVector = _kernel.LoadFunction("SigmoidVector");
_vectorAdd = _kernel.LoadFunction("VectorAdd");
_vectorCopyRandom = _kernel.LoadFunction("VectorCopyRandom");
_copyToMatrix = _kernel.LoadFunction("CopyToMatrix");
_vectorSplit = _kernel.LoadFunction("VectorSplit");
_tensorConvertToVector = _kernel.LoadFunction("TensorConvertToVector");
_tensorConvertToMatrix = _kernel.LoadFunction("TensorConvertToMatrix");
_tensorAddPadding = _kernel.LoadFunction("TensorAddPadding");
_tensorIm2Col = _kernel.LoadFunction("TensorIm2Col");
}