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

Packages that use Tensor

Package	Description
io.github.kirstenali.deepj.activations
io.github.kirstenali.deepj.layers
io.github.kirstenali.deepj.layers.transformer
io.github.kirstenali.deepj.layers.transformer.attention
io.github.kirstenali.deepj.layers.transformer.blocks
io.github.kirstenali.deepj.layers.transformer.norm
io.github.kirstenali.deepj.loss
io.github.kirstenali.deepj.models
io.github.kirstenali.deepj.models.gpt
io.github.kirstenali.deepj.optimisers
io.github.kirstenali.deepj.tensor
io.github.kirstenali.deepj.tensor.cpu
io.github.kirstenali.deepj.tensor.metal
io.github.kirstenali.deepj.training
io.github.kirstenali.deepj.transformer
io.github.kirstenali.deepj.transformer.embeddings

Uses of Tensor in io.github.kirstenali.deepj.activations

Methods in io.github.kirstenali.deepj.activations that return Tensor

Modifier and Type	Method	Description
Tensor	ActivationFunction.backward(Tensor gradOutput)
Tensor	GELU.backward(Tensor gradOutput)
Tensor	ReLU.backward(Tensor gradOutput)
Tensor	Sigmoid.backward(Tensor gradOutput)
Tensor	SiLU.backward(Tensor gradOutput)
Tensor	Softmax.backward(Tensor gradOutput)
Tensor	Tanh.backward(Tensor gradOutput)
Tensor	ActivationFunction.forward(Tensor input)
Tensor	GELU.forward(Tensor input)
Tensor	ReLU.forward(Tensor input)
Tensor	Sigmoid.forward(Tensor input)
Tensor	SiLU.forward(Tensor input)
Tensor	Softmax.forward(Tensor logits)
Tensor	Tanh.forward(Tensor input)

Methods in io.github.kirstenali.deepj.activations with parameters of type Tensor

Modifier and Type	Method	Description
Tensor	ActivationFunction.backward(Tensor gradOutput)
Tensor	GELU.backward(Tensor gradOutput)
Tensor	ReLU.backward(Tensor gradOutput)
Tensor	Sigmoid.backward(Tensor gradOutput)
Tensor	SiLU.backward(Tensor gradOutput)
Tensor	Softmax.backward(Tensor gradOutput)
Tensor	Tanh.backward(Tensor gradOutput)
Tensor	ActivationFunction.forward(Tensor input)
Tensor	GELU.forward(Tensor input)
Tensor	ReLU.forward(Tensor input)
Tensor	Sigmoid.forward(Tensor input)
Tensor	SiLU.forward(Tensor input)
Tensor	Softmax.forward(Tensor logits)
Tensor	Tanh.forward(Tensor input)

Uses of Tensor in io.github.kirstenali.deepj.layers

Methods in io.github.kirstenali.deepj.layers that return Tensor

Modifier and Type	Method	Description
Tensor	FNN.backward(Tensor gradOut)
Tensor	Layer.backward(Tensor gradOutput)
Tensor	Linear.backward(Tensor gradY)
Tensor	FNN.forward(Tensor x)
Tensor	Layer.forward(Tensor input)
Tensor	Linear.forward(Tensor x)

Methods in io.github.kirstenali.deepj.layers with parameters of type Tensor

Modifier and Type	Method	Description
Tensor	FNN.backward(Tensor gradOut)
Tensor	Layer.backward(Tensor gradOutput)
Tensor	Linear.backward(Tensor gradY)
Tensor	FNN.forward(Tensor x)
Tensor	Layer.forward(Tensor input)
Tensor	Linear.forward(Tensor x)

Uses of Tensor in io.github.kirstenali.deepj.layers.transformer

Methods in io.github.kirstenali.deepj.layers.transformer that return Tensor

Modifier and Type	Method	Description
Tensor	SwiGLULayer.backward(Tensor dOut)
Tensor	SwiGLULayer.forward(Tensor x)

Methods in io.github.kirstenali.deepj.layers.transformer with parameters of type Tensor

Modifier and Type	Method	Description
Tensor	SwiGLULayer.backward(Tensor dOut)
Tensor	SwiGLULayer.forward(Tensor x)

Uses of Tensor in io.github.kirstenali.deepj.layers.transformer.attention

Methods in io.github.kirstenali.deepj.layers.transformer.attention that return Tensor

Modifier and Type	Method	Description
Tensor	MultiHeadLatentAttention.backward(Tensor dOut)
Tensor	MultiHeadSelfAttention.backward(Tensor dOut)
Tensor	MultiHeadLatentAttention.forward(Tensor x)
Tensor	MultiHeadSelfAttention.forward(Tensor x)
protected Tensor	MultiHeadSelfAttention.transformQueryKey(Tensor heads, int seqLen)	Hook called on split-head Q and K tensors ([nHeads·seqLen × headDim]) before the scaled dot-product.
protected Tensor	RoPEMultiHeadSelfAttention.transformQueryKey(Tensor heads, int seqLen)	Applies RoPE rotation to Q or K heads in the forward pass.
protected Tensor	MultiHeadSelfAttention.transformQueryKeyBackward(Tensor gradHeads, int seqLen)	Inverse of MultiHeadSelfAttention.transformQueryKey(io.github.kirstenali.deepj.tensor.Tensor, int), called on Q/K gradients in backward.
protected Tensor	RoPEMultiHeadSelfAttention.transformQueryKeyBackward(Tensor gradHeads, int seqLen)	Applies the inverse (transpose) RoPE rotation to Q/K gradients in the backward pass.

Methods in io.github.kirstenali.deepj.layers.transformer.attention with parameters of type Tensor

Modifier and Type	Method	Description
Tensor	MultiHeadLatentAttention.backward(Tensor dOut)
Tensor	MultiHeadSelfAttention.backward(Tensor dOut)
Tensor	MultiHeadLatentAttention.forward(Tensor x)
Tensor	MultiHeadSelfAttention.forward(Tensor x)
protected Tensor	MultiHeadSelfAttention.transformQueryKey(Tensor heads, int seqLen)	Hook called on split-head Q and K tensors ([nHeads·seqLen × headDim]) before the scaled dot-product.
protected Tensor	RoPEMultiHeadSelfAttention.transformQueryKey(Tensor heads, int seqLen)	Applies RoPE rotation to Q or K heads in the forward pass.
protected Tensor	MultiHeadSelfAttention.transformQueryKeyBackward(Tensor gradHeads, int seqLen)	Inverse of MultiHeadSelfAttention.transformQueryKey(io.github.kirstenali.deepj.tensor.Tensor, int), called on Q/K gradients in backward.
protected Tensor	RoPEMultiHeadSelfAttention.transformQueryKeyBackward(Tensor gradHeads, int seqLen)	Applies the inverse (transpose) RoPE rotation to Q/K gradients in the backward pass.

Uses of Tensor in io.github.kirstenali.deepj.layers.transformer.blocks

Methods in io.github.kirstenali.deepj.layers.transformer.blocks that return Tensor

Modifier and Type	Method	Description
Tensor	DeepSeekTransformerBlock.backward(Tensor gradOut)
Tensor	GPTTransformerBlock.backward(Tensor gradOut)
Tensor	LlamaTransformerBlock.backward(Tensor gradOut)
Tensor	DeepSeekTransformerBlock.forward(Tensor x)
Tensor	GPTTransformerBlock.forward(Tensor x)
Tensor	LlamaTransformerBlock.forward(Tensor x)

Methods in io.github.kirstenali.deepj.layers.transformer.blocks with parameters of type Tensor

Modifier and Type	Method	Description
Tensor	DeepSeekTransformerBlock.backward(Tensor gradOut)
Tensor	GPTTransformerBlock.backward(Tensor gradOut)
Tensor	LlamaTransformerBlock.backward(Tensor gradOut)
Tensor	DeepSeekTransformerBlock.forward(Tensor x)
Tensor	GPTTransformerBlock.forward(Tensor x)
Tensor	LlamaTransformerBlock.forward(Tensor x)

Uses of Tensor in io.github.kirstenali.deepj.layers.transformer.norm

Methods in io.github.kirstenali.deepj.layers.transformer.norm that return Tensor

Modifier and Type	Method	Description
Tensor	LayerNorm1D.backward(Tensor gradOut)
Tensor	RMSNorm1D.backward(Tensor gradOut)
Tensor	LayerNorm1D.forward(Tensor x)
Tensor	RMSNorm1D.forward(Tensor x)

Methods in io.github.kirstenali.deepj.layers.transformer.norm with parameters of type Tensor

Modifier and Type	Method	Description
Tensor	LayerNorm1D.backward(Tensor gradOut)
Tensor	RMSNorm1D.backward(Tensor gradOut)
Tensor	LayerNorm1D.forward(Tensor x)
Tensor	RMSNorm1D.forward(Tensor x)

Uses of Tensor in io.github.kirstenali.deepj.loss

Methods in io.github.kirstenali.deepj.loss that return Tensor

Modifier and Type	Method	Description
static Tensor	CrossEntropyLoss.fromIntTargets(int[] targets)	Builds a [n x 1] Tensor from int[] targets.
static Tensor	CrossEntropyLoss.gradient(Tensor logits, int[] targets)	Convenience helper: gradient w.r.t. logits, averaged over rows.
Tensor	CrossEntropyLoss.gradient(Tensor predicted, Tensor actual)
Tensor	LossFunction.gradient(Tensor predicted, Tensor actual)
Tensor	MSELoss.gradient(Tensor predicted, Tensor actual)

Methods in io.github.kirstenali.deepj.loss with parameters of type Tensor

Modifier and Type	Method	Description
static Tensor	CrossEntropyLoss.gradient(Tensor logits, int[] targets)	Convenience helper: gradient w.r.t. logits, averaged over rows.
Tensor	CrossEntropyLoss.gradient(Tensor predicted, Tensor actual)
Tensor	LossFunction.gradient(Tensor predicted, Tensor actual)
Tensor	MSELoss.gradient(Tensor predicted, Tensor actual)
static float	CrossEntropyLoss.loss(Tensor logits, int[] targets)	Convenience helper: compute loss from logits and int targets.
float	CrossEntropyLoss.loss(Tensor predicted, Tensor actual)
float	LossFunction.loss(Tensor predicted, Tensor actual)
float	MSELoss.loss(Tensor predicted, Tensor actual)
static int[]	CrossEntropyLoss.toIntTargets(Tensor actual)	Converts a [n x 1] Tensor of class indices into an int[].

Uses of Tensor in io.github.kirstenali.deepj.models

Methods in io.github.kirstenali.deepj.models that return Tensor

Modifier and Type	Method	Description
protected Tensor	DecoderOnlyModel.embed(int[] inputIds)	Maps input ids to the initial hidden state.
Tensor	CausalLM.forward(int[] inputIds)	Maps token ids to logits [seqLen × vocabSize].
Tensor	DecoderOnlyModel.forward(int[] inputIds)

Methods in io.github.kirstenali.deepj.models with parameters of type Tensor

Modifier and Type	Method	Description
void	CausalLM.backward(Tensor dLogits)	Back-propagates gradient of the logits through the model.
void	DecoderOnlyModel.backward(Tensor dLogits)
protected void	DecoderOnlyModel.backwardEmbeddings(Tensor g)	Back-propagates gradient into embedding layer(s).

Method parameters in io.github.kirstenali.deepj.models with type arguments of type Tensor

Modifier and Type	Method	Description
static String	TextGenerator.generate(Function<int[],Tensor> forwarder, int maxSeqLen, Tokenizer tok, String prompt, int maxNewTokens, float temperature, int topK, long seed)	Generate text with an explicit maxSeqLen — useful when no config is available.
static String	TextGenerator.generate(Function<int[],Tensor> forwarder, TransformerConfig cfg, Tokenizer tok, String prompt, int maxNewTokens, float temperature, int topK, long seed)	Generate text using any model that maps int[] ids → [seqLen × vocabSize] logits.

Uses of Tensor in io.github.kirstenali.deepj.models.gpt

Methods in io.github.kirstenali.deepj.models.gpt that return Tensor

Modifier and Type	Method	Description
protected Tensor	GPTModel.embed(int[] inputIds)

Methods in io.github.kirstenali.deepj.models.gpt with parameters of type Tensor

Modifier and Type	Method	Description
protected void	GPTModel.backwardEmbeddings(Tensor g)

Uses of Tensor in io.github.kirstenali.deepj.optimisers

Fields in io.github.kirstenali.deepj.optimisers declared as Tensor

Modifier and Type	Field	Description
Tensor	Parameter.grad
Tensor	Parameter.value

Constructors in io.github.kirstenali.deepj.optimisers with parameters of type Tensor

Modifier	Constructor	Description
	Parameter(Tensor value)

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

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

Modifier and Type	Method	Description
Tensor	Tensor.add(Tensor other)
Tensor	TensorBackend.add(Tensor a, Tensor b)
Tensor	Tensor.addBroadcastCols(Tensor colVector)
Tensor	TensorBackend.addBroadcastCols(Tensor a, Tensor colVector)
Tensor	Tensor.addBroadcastRows(Tensor rowVector)
Tensor	TensorBackend.addBroadcastRows(Tensor a, Tensor rowVector)
Tensor	Tensor.addInPlace(Tensor b)
Tensor	Tensor.addRowVector(Tensor rowVector)
Tensor	TensorBackend.addRowVector(Tensor a, Tensor rowVector)
Tensor	Tensor.addScalar(float s)
Tensor	TensorBackend.addScalar(Tensor a, float scalar)
Tensor	Tensor.addScalarInPlace(float s)
static Tensor	Tensor.causalMask(int size)
Tensor	Tensor.clamp(float min, float max)
Tensor	TensorBackend.clamp(Tensor a, float min, float max)
Tensor	ComputeGraph.createOutputTensor(GpuBuffer buf)	Create a Tensor backed by a GpuBuffer.
Tensor	Tensor.crossEntropyGradient(int[] targets)
Tensor	TensorBackend.crossEntropyGradient(Tensor logits, int[] targets)
Tensor	Tensor.divide(Tensor other)
Tensor	TensorBackend.divide(Tensor a, Tensor b)
Tensor	Tensor.divideBroadcastCols(Tensor colVector)
Tensor	TensorBackend.divideBroadcastCols(Tensor a, Tensor colVector)
Tensor	Tensor.divideInPlace(Tensor b)
Tensor	Tensor.divideScalar(float s)
Tensor	TensorBackend.divideScalar(Tensor a, float scalar)
Tensor	Tensor.divideScalarInPlace(float s)
Tensor	Tensor.exp()
Tensor	TensorBackend.exp(Tensor a)
Tensor	Tensor.expInPlace()
static Tensor	Tensor.from2D(float[][] data)	Build a tensor from 2-D row-major data.
static Tensor	TensorAdapters.fromIntColumn(int[] values)	Build a [n x 1] tensor from integer ids (stored as float values).
Tensor	TensorBackend.gelu(Tensor a)
Tensor	Tensor.geluActivation()
Tensor	Tensor.geluBackward(Tensor gradOutput)
Tensor	TensorBackend.geluBackward(Tensor input, Tensor gradOutput)
Tensor	Tensor.geluInPlace()
Tensor	Tensor.getRow(int row)
static Tensor	Tensor.layerNormBackward(Tensor dXHat, Tensor xHat, Tensor std, int dim)
Tensor	TensorBackend.layerNormBackward(Tensor dXHat, Tensor xHat, Tensor std, int dim)	LayerNorm backward through normalization (given dXHat, xHat, std).
Tensor	Tensor.log()
Tensor	TensorBackend.log(Tensor a)
Tensor	Tensor.logInPlace()
Tensor	Tensor.matmul(Tensor other)
Tensor	TensorBackend.matmul(Tensor a, Tensor b)
Tensor	Tensor.maxAlongRows()
Tensor	TensorBackend.maxAlongRows(Tensor a)
Tensor	Tensor.meanAlongRows()
Tensor	TensorBackend.meanAlongRows(Tensor a)
Tensor	Tensor.multiply(Tensor other)
Tensor	TensorBackend.multiply(Tensor a, Tensor b)
Tensor	Tensor.multiplyBroadcastCols(Tensor colVector)
Tensor	TensorBackend.multiplyBroadcastCols(Tensor a, Tensor colVector)
Tensor	Tensor.multiplyBroadcastRows(Tensor rowVector)
Tensor	TensorBackend.multiplyBroadcastRows(Tensor a, Tensor rowVector)
Tensor	Tensor.multiplyInPlace(Tensor b)
Tensor	Tensor.multiplyScalar(float s)
Tensor	TensorBackend.multiplyScalar(Tensor a, float scalar)
Tensor	Tensor.multiplyScalarInPlace(float s)
Tensor	Tensor.neg()
Tensor	TensorBackend.neg(Tensor a)
Tensor	Tensor.negInPlace()
static Tensor	Tensor.ones(int rows, int cols)
Tensor	Tensor.pow(float exponent)
Tensor	TensorBackend.pow(Tensor a, float exponent)
static Tensor	Tensor.random(int rows, int cols, Random rand)
Tensor	TensorBackend.relu(Tensor a)
Tensor	Tensor.reluActivation()
Tensor	Tensor.reluBackward(Tensor gradOutput)
Tensor	TensorBackend.reluBackward(Tensor input, Tensor gradOutput)
Tensor	Tensor.reluInPlace()
static Tensor	Tensor.sampleRows(Tensor t, int n, Random rnd)
Tensor	TensorBackend.sigmoid(Tensor a)
Tensor	Tensor.sigmoidActivation()
Tensor	Tensor.sigmoidInPlace()
static Tensor	Tensor.sliceRows(Tensor t, int[] rowIndices, int cols)
Tensor	Tensor.softmaxBackward(Tensor softmaxOut)
Tensor	TensorBackend.softmaxBackward(Tensor gradOutput, Tensor softmaxOut)
Tensor	Tensor.softmaxRows()
Tensor	TensorBackend.softmaxRows(Tensor logits)
Tensor	Tensor.sqrt()
Tensor	TensorBackend.sqrt(Tensor a)
Tensor	Tensor.sqrtInPlace()
Tensor	Tensor.subtract(Tensor other)
Tensor	TensorBackend.subtract(Tensor a, Tensor b)
Tensor	Tensor.subtractBroadcastCols(Tensor colVector)
Tensor	TensorBackend.subtractBroadcastCols(Tensor a, Tensor colVector)
Tensor	Tensor.subtractInPlace(Tensor b)
Tensor	Tensor.sumAlongCols()
Tensor	TensorBackend.sumAlongCols(Tensor a)
Tensor	Tensor.sumAlongRows()
Tensor	TensorBackend.sumAlongRows(Tensor a)
Tensor	Tensor.sumRows()
Tensor	TensorBackend.sumRows(Tensor a)
Tensor	TensorBackend.tanh(Tensor a)
Tensor	Tensor.tanhActivation()
Tensor	Tensor.tanhInPlace()
Tensor	Tensor.transpose()
Tensor	TensorBackend.transpose(Tensor a)
static Tensor	TensorAdapters.unpackF32(float[] flat, int rows, int cols)	Unpack a flat float32 array into a new Tensor.
Tensor	Tensor.varianceAlongRows()
Tensor	TensorBackend.varianceAlongRows(Tensor a)
static Tensor	Tensor.zeros(int rows, int cols)

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

Modifier and Type	Method	Description
static void	Tensor.adamWUpdate(Tensor w, Tensor g, Tensor mt, Tensor vt, float lr, float beta1, float beta2, float eps, float weightDecay, float bc1, float bc2)
void	TensorBackend.adamWUpdate(Tensor w, Tensor g, Tensor mt, Tensor vt, float lr, float beta1, float beta2, float eps, float weightDecay, float bc1, float bc2)	In-place AdamW update.
Tensor	Tensor.add(Tensor other)
Tensor	TensorBackend.add(Tensor a, Tensor b)
Tensor	Tensor.addBroadcastCols(Tensor colVector)
Tensor	TensorBackend.addBroadcastCols(Tensor a, Tensor colVector)
Tensor	Tensor.addBroadcastRows(Tensor rowVector)
Tensor	TensorBackend.addBroadcastRows(Tensor a, Tensor rowVector)
Tensor	Tensor.addInPlace(Tensor b)
void	TensorBackend.addInPlace(Tensor a, Tensor b)
Tensor	Tensor.addRowVector(Tensor rowVector)
Tensor	TensorBackend.addRowVector(Tensor a, Tensor rowVector)
Tensor	TensorBackend.addScalar(Tensor a, float scalar)
void	TensorBackend.addScalarInPlace(Tensor a, float s)
void	ComputeGraph.bindTensorToBuffer(Tensor t, GpuBuffer buf)	Rebind an existing tensor to a GPU buffer and track ownership for lifecycle management.
Tensor	TensorBackend.clamp(Tensor a, float min, float max)
Tensor	TensorBackend.crossEntropyGradient(Tensor logits, int[] targets)
float	TensorBackend.crossEntropyLoss(Tensor logits, int[] targets)
Tensor	Tensor.divide(Tensor other)
Tensor	TensorBackend.divide(Tensor a, Tensor b)
Tensor	Tensor.divideBroadcastCols(Tensor colVector)
Tensor	TensorBackend.divideBroadcastCols(Tensor a, Tensor colVector)
Tensor	Tensor.divideInPlace(Tensor b)
void	TensorBackend.divideInPlace(Tensor a, Tensor b)
Tensor	TensorBackend.divideScalar(Tensor a, float scalar)
void	TensorBackend.divideScalarInPlace(Tensor a, float s)
GpuBuffer	ComputeGraph.ensureGpuBuffer(Tensor t)	Ensure a tensor has a GpuBuffer.
Tensor	TensorBackend.exp(Tensor a)
void	TensorBackend.expInPlace(Tensor a)
Tensor	TensorBackend.gelu(Tensor a)
Tensor	Tensor.geluBackward(Tensor gradOutput)
Tensor	TensorBackend.geluBackward(Tensor input, Tensor gradOutput)
void	TensorBackend.geluInPlace(Tensor a)
static Tensor	Tensor.layerNormBackward(Tensor dXHat, Tensor xHat, Tensor std, int dim)
Tensor	TensorBackend.layerNormBackward(Tensor dXHat, Tensor xHat, Tensor std, int dim)	LayerNorm backward through normalization (given dXHat, xHat, std).
Tensor	TensorBackend.log(Tensor a)
void	TensorBackend.logInPlace(Tensor a)
void	ComputeGraph.materialize(Tensor t)	Materialize a tensor: flush pending ops if needed, then download GPU data to CPU.
default void	TensorBackend.materializeTensor(Tensor t)	Materialize a tensor: flush any pending GPU computation and download the result to the tensor's CPU data[].
Tensor	Tensor.matmul(Tensor other)
Tensor	TensorBackend.matmul(Tensor a, Tensor b)
Tensor	TensorBackend.maxAlongRows(Tensor a)
Tensor	TensorBackend.meanAlongRows(Tensor a)
Tensor	Tensor.multiply(Tensor other)
Tensor	TensorBackend.multiply(Tensor a, Tensor b)
Tensor	Tensor.multiplyBroadcastCols(Tensor colVector)
Tensor	TensorBackend.multiplyBroadcastCols(Tensor a, Tensor colVector)
Tensor	Tensor.multiplyBroadcastRows(Tensor rowVector)
Tensor	TensorBackend.multiplyBroadcastRows(Tensor a, Tensor rowVector)
Tensor	Tensor.multiplyInPlace(Tensor b)
void	TensorBackend.multiplyInPlace(Tensor a, Tensor b)
Tensor	TensorBackend.multiplyScalar(Tensor a, float scalar)
void	TensorBackend.multiplyScalarInPlace(Tensor a, float s)
Tensor	TensorBackend.neg(Tensor a)
void	TensorBackend.negInPlace(Tensor a)
static float[]	TensorAdapters.packF32(Tensor t)	Pack a Tensor's data into a flat float32 array (row-major).
Tensor	TensorBackend.pow(Tensor a, float exponent)
Tensor	TensorBackend.relu(Tensor a)
Tensor	Tensor.reluBackward(Tensor gradOutput)
Tensor	TensorBackend.reluBackward(Tensor input, Tensor gradOutput)
void	TensorBackend.reluInPlace(Tensor a)
static void	Tensor.requireSameShape(Tensor a, Tensor b, String op)
static void	Tensor.requireTargetsMatchRows(Tensor logits, int[] targets)
static Tensor	Tensor.sampleRows(Tensor t, int n, Random rnd)
static void	Tensor.scatterAddRows(Tensor target, int[] indices, Tensor grad)
void	TensorBackend.scatterAddRows(Tensor target, int[] indices, Tensor grad)
void	Tensor.setRow(int row, Tensor source, int srcRow)
Tensor	TensorBackend.sigmoid(Tensor a)
void	TensorBackend.sigmoidInPlace(Tensor a)
static Tensor	Tensor.sliceRows(Tensor t, int[] rowIndices, int cols)
Tensor	Tensor.softmaxBackward(Tensor softmaxOut)
Tensor	TensorBackend.softmaxBackward(Tensor gradOutput, Tensor softmaxOut)
Tensor	TensorBackend.softmaxRows(Tensor logits)
Tensor	TensorBackend.sqrt(Tensor a)
void	TensorBackend.sqrtInPlace(Tensor a)
Tensor	Tensor.subtract(Tensor other)
Tensor	TensorBackend.subtract(Tensor a, Tensor b)
Tensor	Tensor.subtractBroadcastCols(Tensor colVector)
Tensor	TensorBackend.subtractBroadcastCols(Tensor a, Tensor colVector)
Tensor	Tensor.subtractInPlace(Tensor b)
void	TensorBackend.subtractInPlace(Tensor a, Tensor b)
float	TensorBackend.sum(Tensor a)
float	TensorBackend.sumAbs(Tensor a)
Tensor	TensorBackend.sumAlongCols(Tensor a)
Tensor	TensorBackend.sumAlongRows(Tensor a)
Tensor	TensorBackend.sumRows(Tensor a)
Tensor	TensorBackend.tanh(Tensor a)
void	TensorBackend.tanhInPlace(Tensor a)
Tensor	TensorBackend.transpose(Tensor a)
static void	TensorAdapters.unpackF32Into(float[] flat, Tensor t)	Unpack a flat float32 array into an existing tensor's data[].
Tensor	TensorBackend.varianceAlongRows(Tensor a)

Constructors in io.github.kirstenali.deepj.tensor with parameters of type Tensor

Modifier	Constructor	Description
	Tensor(Tensor source)	Copy constructor — creates an independent deep copy.

Uses of Tensor in io.github.kirstenali.deepj.tensor.cpu

Methods in io.github.kirstenali.deepj.tensor.cpu that return Tensor

Modifier and Type	Method	Description
Tensor	CpuBackend.add(Tensor a, Tensor b)
Tensor	CpuBackend.addBroadcastCols(Tensor a, Tensor cv)
Tensor	CpuBackend.addBroadcastRows(Tensor a, Tensor rv)
Tensor	CpuBackend.addRowVector(Tensor a, Tensor rv)
Tensor	CpuBackend.addScalar(Tensor a, float s)
Tensor	CpuBackend.causalMask(int size)
Tensor	CpuBackend.clamp(Tensor a, float min, float max)
Tensor	CpuBackend.crossEntropyGradient(Tensor logits, int[] targets)
Tensor	CpuBackend.divide(Tensor a, Tensor b)
Tensor	CpuBackend.divideBroadcastCols(Tensor a, Tensor cv)
Tensor	CpuBackend.divideScalar(Tensor a, float s)
Tensor	CpuBackend.exp(Tensor a)
Tensor	CpuBackend.gelu(Tensor a)
Tensor	CpuBackend.geluBackward(Tensor input, Tensor gradOutput)
Tensor	CpuBackend.getRow(Tensor t, int row)
Tensor	CpuBackend.layerNormBackward(Tensor dXHat, Tensor xHat, Tensor std, int dim)
Tensor	CpuBackend.log(Tensor a)
Tensor	CpuBackend.matmul(Tensor a, Tensor b)
Tensor	CpuBackend.maxAlongRows(Tensor a)
Tensor	CpuBackend.meanAlongRows(Tensor a)
Tensor	CpuBackend.multiply(Tensor a, Tensor b)
Tensor	CpuBackend.multiplyBroadcastCols(Tensor a, Tensor cv)
Tensor	CpuBackend.multiplyBroadcastRows(Tensor a, Tensor rv)
Tensor	CpuBackend.multiplyScalar(Tensor a, float s)
Tensor	CpuBackend.neg(Tensor a)
Tensor	CpuBackend.ones(int rows, int cols)
Tensor	CpuBackend.pow(Tensor a, float exponent)
Tensor	CpuBackend.random(int rows, int cols, Random rand)
Tensor	CpuBackend.relu(Tensor a)
Tensor	CpuBackend.reluBackward(Tensor input, Tensor gradOutput)
Tensor	CpuBackend.sampleRows(Tensor t, int n, Random rnd)
Tensor	CpuBackend.sigmoid(Tensor a)
Tensor	CpuBackend.sliceRows(Tensor t, int[] rowIndices, int cols)
Tensor	CpuBackend.softmaxBackward(Tensor gradOutput, Tensor softmaxOut)
Tensor	CpuBackend.softmaxRows(Tensor logits)
Tensor	CpuBackend.sqrt(Tensor a)
Tensor	CpuBackend.subtract(Tensor a, Tensor b)
Tensor	CpuBackend.subtractBroadcastCols(Tensor a, Tensor cv)
Tensor	CpuBackend.sumAlongCols(Tensor a)
Tensor	CpuBackend.sumAlongRows(Tensor a)
Tensor	CpuBackend.sumRows(Tensor a)
Tensor	CpuBackend.tanh(Tensor a)
Tensor	CpuBackend.transpose(Tensor a)
Tensor	CpuBackend.varianceAlongRows(Tensor a)
Tensor	CpuBackend.zeros(int rows, int cols)

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

Modifier and Type	Method	Description
void	CpuBackend.adamWUpdate(Tensor w, Tensor g, Tensor mt, Tensor vt, float lr, float beta1, float beta2, float eps, float weightDecay, float bc1, float bc2)
Tensor	CpuBackend.add(Tensor a, Tensor b)
Tensor	CpuBackend.addBroadcastCols(Tensor a, Tensor cv)
Tensor	CpuBackend.addBroadcastRows(Tensor a, Tensor rv)
void	CpuBackend.addInPlace(Tensor a, Tensor b)
Tensor	CpuBackend.addRowVector(Tensor a, Tensor rv)
Tensor	CpuBackend.addScalar(Tensor a, float s)
void	CpuBackend.addScalarInPlace(Tensor a, float s)
Tensor	CpuBackend.clamp(Tensor a, float min, float max)
Tensor	CpuBackend.crossEntropyGradient(Tensor logits, int[] targets)
float	CpuBackend.crossEntropyLoss(Tensor logits, int[] targets)
Tensor	CpuBackend.divide(Tensor a, Tensor b)
Tensor	CpuBackend.divideBroadcastCols(Tensor a, Tensor cv)
void	CpuBackend.divideInPlace(Tensor a, Tensor b)
Tensor	CpuBackend.divideScalar(Tensor a, float s)
void	CpuBackend.divideScalarInPlace(Tensor a, float s)
Tensor	CpuBackend.exp(Tensor a)
void	CpuBackend.expInPlace(Tensor a)
Tensor	CpuBackend.gelu(Tensor a)
Tensor	CpuBackend.geluBackward(Tensor input, Tensor gradOutput)
void	CpuBackend.geluInPlace(Tensor a)
float	CpuBackend.get(Tensor t, int r, int c)
Tensor	CpuBackend.getRow(Tensor t, int row)
Tensor	CpuBackend.layerNormBackward(Tensor dXHat, Tensor xHat, Tensor std, int dim)
Tensor	CpuBackend.log(Tensor a)
void	CpuBackend.logInPlace(Tensor a)
Tensor	CpuBackend.matmul(Tensor a, Tensor b)
Tensor	CpuBackend.maxAlongRows(Tensor a)
Tensor	CpuBackend.meanAlongRows(Tensor a)
Tensor	CpuBackend.multiply(Tensor a, Tensor b)
Tensor	CpuBackend.multiplyBroadcastCols(Tensor a, Tensor cv)
Tensor	CpuBackend.multiplyBroadcastRows(Tensor a, Tensor rv)
void	CpuBackend.multiplyInPlace(Tensor a, Tensor b)
Tensor	CpuBackend.multiplyScalar(Tensor a, float s)
void	CpuBackend.multiplyScalarInPlace(Tensor a, float s)
Tensor	CpuBackend.neg(Tensor a)
void	CpuBackend.negInPlace(Tensor a)
Tensor	CpuBackend.pow(Tensor a, float exponent)
void	CpuBackend.print(Tensor t, String label)
Tensor	CpuBackend.relu(Tensor a)
Tensor	CpuBackend.reluBackward(Tensor input, Tensor gradOutput)
void	CpuBackend.reluInPlace(Tensor a)
Tensor	CpuBackend.sampleRows(Tensor t, int n, Random rnd)
void	CpuBackend.scatterAddRows(Tensor target, int[] indices, Tensor grad)
void	CpuBackend.set(Tensor t, int r, int c, float v)
void	CpuBackend.setRow(Tensor t, int row, Tensor source, int srcRow)
Tensor	CpuBackend.sigmoid(Tensor a)
void	CpuBackend.sigmoidInPlace(Tensor a)
Tensor	CpuBackend.sliceRows(Tensor t, int[] rowIndices, int cols)
Tensor	CpuBackend.softmaxBackward(Tensor gradOutput, Tensor softmaxOut)
Tensor	CpuBackend.softmaxRows(Tensor logits)
Tensor	CpuBackend.sqrt(Tensor a)
void	CpuBackend.sqrtInPlace(Tensor a)
Tensor	CpuBackend.subtract(Tensor a, Tensor b)
Tensor	CpuBackend.subtractBroadcastCols(Tensor a, Tensor cv)
void	CpuBackend.subtractInPlace(Tensor a, Tensor b)
float	CpuBackend.sum(Tensor a)
float	CpuBackend.sumAbs(Tensor a)
Tensor	CpuBackend.sumAlongCols(Tensor a)
Tensor	CpuBackend.sumAlongRows(Tensor a)
Tensor	CpuBackend.sumRows(Tensor a)
Tensor	CpuBackend.tanh(Tensor a)
void	CpuBackend.tanhInPlace(Tensor a)
Tensor	CpuBackend.transpose(Tensor a)
Tensor	CpuBackend.varianceAlongRows(Tensor a)

Uses of Tensor in io.github.kirstenali.deepj.tensor.metal

Methods in io.github.kirstenali.deepj.tensor.metal that return Tensor

Modifier and Type	Method	Description
Tensor	MetalBackend.add(Tensor a, Tensor b)
Tensor	MetalBackend.addBroadcastCols(Tensor a, Tensor v)
Tensor	MetalBackend.addBroadcastRows(Tensor a, Tensor v)
Tensor	MetalBackend.addRowVector(Tensor a, Tensor v)
Tensor	MetalBackend.addScalar(Tensor a, float scalar)
Tensor	MetalBackend.clamp(Tensor a, float min, float max)
Tensor	MetalBackend.crossEntropyGradient(Tensor logits, int[] targets)
Tensor	MetalBackend.divide(Tensor a, Tensor b)
Tensor	MetalBackend.divideBroadcastCols(Tensor a, Tensor v)
Tensor	MetalBackend.divideScalar(Tensor a, float scalar)
Tensor	MetalBackend.exp(Tensor a)
Tensor	MetalBackend.gelu(Tensor a)
Tensor	MetalBackend.geluBackward(Tensor input, Tensor gradOutput)
Tensor	MetalBackend.layerNormBackward(Tensor dXHat, Tensor xHat, Tensor std, int dim)
Tensor	MetalBackend.log(Tensor a)
Tensor	MetalBackend.matmul(Tensor a, Tensor b)
Tensor	MetalBackend.maxAlongRows(Tensor a)
Tensor	MetalBackend.meanAlongRows(Tensor a)
Tensor	MetalBackend.multiply(Tensor a, Tensor b)
Tensor	MetalBackend.multiplyBroadcastCols(Tensor a, Tensor v)
Tensor	MetalBackend.multiplyBroadcastRows(Tensor a, Tensor v)
Tensor	MetalBackend.multiplyScalar(Tensor a, float scalar)
Tensor	MetalBackend.neg(Tensor a)
Tensor	MetalBackend.pow(Tensor a, float exponent)
Tensor	MetalBackend.relu(Tensor a)
Tensor	MetalBackend.reluBackward(Tensor input, Tensor gradOutput)
Tensor	MetalBackend.sigmoid(Tensor a)
Tensor	MetalBackend.softmaxBackward(Tensor gradOutput, Tensor softmaxOut)
Tensor	MetalBackend.softmaxRows(Tensor logits)
Tensor	MetalBackend.sqrt(Tensor a)
Tensor	MetalBackend.subtract(Tensor a, Tensor b)
Tensor	MetalBackend.subtractBroadcastCols(Tensor a, Tensor v)
Tensor	MetalBackend.sumAlongCols(Tensor a)
Tensor	MetalBackend.sumAlongRows(Tensor a)
Tensor	MetalBackend.sumRows(Tensor a)
Tensor	MetalBackend.tanh(Tensor a)
Tensor	MetalBackend.transpose(Tensor a)
Tensor	MetalBackend.varianceAlongRows(Tensor a)

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

Modifier and Type	Method	Description
void	MetalBackend.adamWUpdate(Tensor w, Tensor g, Tensor mt, Tensor vt, float lr, float beta1, float beta2, float eps, float weightDecay, float bc1, float bc2)
Tensor	MetalBackend.add(Tensor a, Tensor b)
Tensor	MetalBackend.addBroadcastCols(Tensor a, Tensor v)
Tensor	MetalBackend.addBroadcastRows(Tensor a, Tensor v)
void	MetalBackend.addInPlace(Tensor a, Tensor b)
Tensor	MetalBackend.addRowVector(Tensor a, Tensor v)
Tensor	MetalBackend.addScalar(Tensor a, float scalar)
void	MetalBackend.addScalarInPlace(Tensor a, float s)
Tensor	MetalBackend.clamp(Tensor a, float min, float max)
Tensor	MetalBackend.crossEntropyGradient(Tensor logits, int[] targets)
float	MetalBackend.crossEntropyLoss(Tensor logits, int[] targets)
Tensor	MetalBackend.divide(Tensor a, Tensor b)
Tensor	MetalBackend.divideBroadcastCols(Tensor a, Tensor v)
void	MetalBackend.divideInPlace(Tensor a, Tensor b)
Tensor	MetalBackend.divideScalar(Tensor a, float scalar)
void	MetalBackend.divideScalarInPlace(Tensor a, float s)
Tensor	MetalBackend.exp(Tensor a)
void	MetalBackend.expInPlace(Tensor a)
Tensor	MetalBackend.gelu(Tensor a)
Tensor	MetalBackend.geluBackward(Tensor input, Tensor gradOutput)
void	MetalBackend.geluInPlace(Tensor a)
Tensor	MetalBackend.layerNormBackward(Tensor dXHat, Tensor xHat, Tensor std, int dim)
Tensor	MetalBackend.log(Tensor a)
void	MetalBackend.logInPlace(Tensor a)
void	MetalBackend.materializeTensor(Tensor t)
Tensor	MetalBackend.matmul(Tensor a, Tensor b)
Tensor	MetalBackend.maxAlongRows(Tensor a)
Tensor	MetalBackend.meanAlongRows(Tensor a)
Tensor	MetalBackend.multiply(Tensor a, Tensor b)
Tensor	MetalBackend.multiplyBroadcastCols(Tensor a, Tensor v)
Tensor	MetalBackend.multiplyBroadcastRows(Tensor a, Tensor v)
void	MetalBackend.multiplyInPlace(Tensor a, Tensor b)
Tensor	MetalBackend.multiplyScalar(Tensor a, float scalar)
void	MetalBackend.multiplyScalarInPlace(Tensor a, float s)
Tensor	MetalBackend.neg(Tensor a)
void	MetalBackend.negInPlace(Tensor a)
Tensor	MetalBackend.pow(Tensor a, float exponent)
Tensor	MetalBackend.relu(Tensor a)
Tensor	MetalBackend.reluBackward(Tensor input, Tensor gradOutput)
void	MetalBackend.reluInPlace(Tensor a)
void	MetalBackend.scatterAddRows(Tensor target, int[] indices, Tensor grad)
Tensor	MetalBackend.sigmoid(Tensor a)
void	MetalBackend.sigmoidInPlace(Tensor a)
Tensor	MetalBackend.softmaxBackward(Tensor gradOutput, Tensor softmaxOut)
Tensor	MetalBackend.softmaxRows(Tensor logits)
Tensor	MetalBackend.sqrt(Tensor a)
void	MetalBackend.sqrtInPlace(Tensor a)
Tensor	MetalBackend.subtract(Tensor a, Tensor b)
Tensor	MetalBackend.subtractBroadcastCols(Tensor a, Tensor v)
void	MetalBackend.subtractInPlace(Tensor a, Tensor b)
float	MetalBackend.sum(Tensor a)
float	MetalBackend.sumAbs(Tensor a)
Tensor	MetalBackend.sumAlongCols(Tensor a)
Tensor	MetalBackend.sumAlongRows(Tensor a)
Tensor	MetalBackend.sumRows(Tensor a)
Tensor	MetalBackend.tanh(Tensor a)
void	MetalBackend.tanhInPlace(Tensor a)
Tensor	MetalBackend.transpose(Tensor a)
Tensor	MetalBackend.varianceAlongRows(Tensor a)

Uses of Tensor in io.github.kirstenali.deepj.training

Methods in io.github.kirstenali.deepj.training with parameters of type Tensor

Modifier and Type	Method	Description
static Trainer	SupervisedTraining.trainer(Layer model, LossFunction lossFn, ParameterOptimizer opt, Tensor xAll, Tensor yAll, long seed)

Uses of Tensor in io.github.kirstenali.deepj.transformer

Methods in io.github.kirstenali.deepj.transformer that return Tensor

Modifier and Type	Method	Description
Tensor	TransformerStack.backward(Tensor gradOut)
Tensor	TransformerStack.forward(Tensor x)

Methods in io.github.kirstenali.deepj.transformer with parameters of type Tensor

Modifier and Type	Method	Description
Tensor	TransformerStack.backward(Tensor gradOut)
Tensor	TransformerStack.forward(Tensor x)

Uses of Tensor in io.github.kirstenali.deepj.transformer.embeddings

Methods in io.github.kirstenali.deepj.transformer.embeddings that return Tensor

Modifier and Type	Method	Description
Tensor	RotaryEmbedding.apply(Tensor t, int seqLen, int nHeads)	Apply rotary embeddings to a split-head tensor (forward direction).
Tensor	RotaryEmbedding.applyBackward(Tensor t, int seqLen, int nHeads)	Apply the transpose (inverse) rotation — used in the backward pass.
Tensor	Embedding.forward(int[] ids)
Tensor	PositionalEmbedding.forward(int seqLen)

Methods in io.github.kirstenali.deepj.transformer.embeddings with parameters of type Tensor

Modifier and Type	Method	Description
Tensor	RotaryEmbedding.apply(Tensor t, int seqLen, int nHeads)	Apply rotary embeddings to a split-head tensor (forward direction).
Tensor	RotaryEmbedding.applyBackward(Tensor t, int seqLen, int nHeads)	Apply the transpose (inverse) rotation — used in the backward pass.
void	Embedding.backward(Tensor gradOut)
void	PositionalEmbedding.backward(Tensor gradOut)