Uses of Class
io.github.kirstenali.deepj.tensor.GpuBuffer

Packages that use GpuBuffer

Package	Description
io.github.kirstenali.deepj.tensor

Uses of GpuBuffer in io.github.kirstenali.deepj.tensor

Methods in io.github.kirstenali.deepj.tensor that return GpuBuffer

Modifier and Type	Method	Description
GpuBuffer	ComputeGraph.ensureGpuBuffer(Tensor t)	Ensure a tensor has a GpuBuffer.
GpuBuffer	ComputeGraph.newOutputBuffer(int rows, int cols)	Allocate a new output buffer (result of a GPU op).

Methods in io.github.kirstenali.deepj.tensor with parameters of type GpuBuffer

Modifier and Type	Method	Description
void	ComputeGraph.bindTensorToBuffer(Tensor t, GpuBuffer buf)	Rebind an existing tensor to a GPU buffer and track ownership for lifecycle management.
Tensor	ComputeGraph.createOutputTensor(GpuBuffer buf)	Create a Tensor backed by a GpuBuffer.
void	ComputeGraph.recordAdamWUpdate(GpuBuffer w, GpuBuffer g, GpuBuffer mt, GpuBuffer vt, float lr, float beta1, float beta2, float eps, float weightDecay, float bc1, float bc2, int n)	Record in-place AdamW update: [OP_ADAMW_UPDATE, wId, gId, mtId, vtId, lrBits, beta1Bits, beta2Bits, epsBits, weightDecayBits, bc1Bits, bc2Bits, n]
void	ComputeGraph.recordBinary(int opCode, GpuBuffer a, GpuBuffer b, GpuBuffer out)	Record a binary element-wise op: [opCode, aId, bId, outId, n]
void	ComputeGraph.recordClamp(GpuBuffer in, GpuBuffer out, float min, float max)	Record clamp: [OP_CLAMP, inId, outId, minBits, maxBits, n]
void	ComputeGraph.recordColBroadcast(int opCode, GpuBuffer a, GpuBuffer colVec, GpuBuffer out, int rows, int cols)	Record col broadcast: [opCode, aId, colVecId, outId, rows, cols]
void	ComputeGraph.recordCrossEntropyGradient(GpuBuffer logits, GpuBuffer targets, GpuBuffer out, int rows, int cols)	Record cross-entropy gradient: [OP_CROSS_ENTROPY_GRADIENT, logitsId, targetsId, outId, rows, cols]
void	ComputeGraph.recordCrossEntropyLoss(GpuBuffer logits, GpuBuffer targets, GpuBuffer out, int rows, int cols)	Record cross-entropy row losses: [OP_CROSS_ENTROPY_LOSS, logitsId, targetsId, outId, rows, cols]
void	ComputeGraph.recordLayerNormBackward(GpuBuffer dXHat, GpuBuffer xHat, GpuBuffer std, GpuBuffer out, int rows, int cols)	Record layer norm backward: [OP_LAYERNORM_BACKWARD, dXHatId, xHatId, stdId, outId, rows, cols]
void	ComputeGraph.recordMatmul(GpuBuffer a, GpuBuffer b, GpuBuffer out, int m, int n, int k)	Record matmul: [OP_MATMUL, aId, bId, outId, m, n, k]
void	ComputeGraph.recordMultiplyScalar(GpuBuffer in, GpuBuffer out, float scalar)	Record scalar multiply: [OP_MULTIPLY_SCALAR, inId, outId, scalarBits, n]
void	ComputeGraph.recordPow(GpuBuffer in, GpuBuffer out, float exponent)	Record pow: [OP_POW, inId, outId, exponentBits, n]
void	ComputeGraph.recordReduction(int opCode, GpuBuffer in, GpuBuffer out, int rows, int cols)	Record row/col reduction: [opCode, inId, outId, rows, cols]
void	ComputeGraph.recordRowBroadcast(int opCode, GpuBuffer a, GpuBuffer rowVec, GpuBuffer out, int rows, int cols)	Record row broadcast: [opCode, aId, rowVecId, outId, rows, cols]
void	ComputeGraph.recordScalarUnary(int opCode, GpuBuffer in, GpuBuffer out, float scalar)	Record scalar add/divide: [opCode, inId, outId, scalarBits, n]
void	ComputeGraph.recordScatterAddRows(GpuBuffer target, GpuBuffer indices, GpuBuffer grad, int targetRows, int targetCols, int nIndices)	Record scatter-add-rows: [OP_SCATTER_ADD_ROWS, targetId, indicesId, gradId, targetRows, targetCols, nIdx]
void	ComputeGraph.recordScatterAddRowsAtomic(GpuBuffer target, GpuBuffer indices, GpuBuffer grad, int targetRows, int targetCols, int nIndices)	Record scatter-add-rows (atomic): [OP_SCATTER_ADD_ROWS_ATOMIC, targetId, indicesId, gradId, targetRows, targetCols, nIdx]
void	ComputeGraph.recordSoftmaxBackward(GpuBuffer gradOutput, GpuBuffer softmaxOut, GpuBuffer out, int rows, int cols)	Record softmax backward: [OP_SOFTMAX_BACKWARD, gradId, softmaxId, outId, rows, cols]
void	ComputeGraph.recordSoftmaxRows(GpuBuffer in, GpuBuffer out, int rows, int cols)	Record softmax rows: [OP_SOFTMAX_ROWS, inId, outId, rows, cols]
void	ComputeGraph.recordSumAbs(GpuBuffer in, GpuBuffer out, int rows, int cols)	Record sum-abs row reduction: [OP_SUM_ABS, inId, outId, rows, cols]
void	ComputeGraph.recordSumScalar(GpuBuffer in, GpuBuffer out, int rows, int cols)	Record scalar sum reduction: [OP_SUM_SCALAR, inId, outId, rows, cols]
void	ComputeGraph.recordTranspose(GpuBuffer in, GpuBuffer out, int rows, int cols)	Record transpose: [OP_TRANSPOSE, inId, outId, rows, cols]
void	ComputeGraph.recordUnary(int opCode, GpuBuffer in, GpuBuffer out)	Record a unary op: [opCode, inId, outId, n]