Tensor class
Constructors
-
Tensor.fill(List<
int> shape, double val) -
factory
-
Tensor.fromList(List<
int> shape, List<double> vals) -
factory
Properties
-
data
↔ List<
double> -
getter/setter pair
-
grad
→ List<
double> -
no setter
-
handle
→ Pointer<
Void> -
no setter
- hashCode → int
-
The hash code for this object.
no setterinherited
- isView ↔ bool?
-
getter/setter pair
- length ↔ int
-
latefinal
- runtimeType → Type
-
A representation of the runtime type of the object.
no setterinherited
-
shape
↔ List<
int> -
latefinal
Methods
-
abs(
) → Tensor -
backward(
) → void -
computeCostMatrix(
Tensor gtBoxes) → Tensor -
computeCrossEntropy(
Tensor pred, List< double> targets, int vocabSize, List<Tensor> tracker) → Tensor -
crossEntropy(
List< int> targets) → Tensor -
detach(
) → Tensor -
dispose(
) → void -
fetchData(
) → Float32List -
fetchRow(
int row) → List< double> -
gelu(
) → Tensor -
getRow(
int row) → Tensor -
log(
) → Tensor -
matmul(
Tensor o) → Tensor -
Matrix Multiplication (Dot Product)
Result shape:
M, N -
mean(
) → Tensor - Collapses the entire tensor into a 1x1 Tensor containing the mean
-
mseLoss(
Tensor target) → Tensor -
normalize(
{double eps = 1e-12}) → Tensor -
noSuchMethod(
Invocation invocation) → dynamic -
Invoked when a nonexistent method or property is accessed.
inherited
-
pow(
double e) → Tensor -
printGradient(
) → String -
printMatrix(
) → String -
relu(
) → Tensor -
reshape(
List< int> newShape) → Tensor -
sigmoid(
) → Tensor -
slice(
int startRow, int rowCount) → Tensor -
softmax(
) → Tensor -
step(
double lr) → void -
sum(
) → Tensor -
toString(
) → String -
A string representation of this object.
inherited
-
zeroGrad(
) → void
Operators
-
operator *(
dynamic o) → Tensor - Element-wise Multiplication (Hadamard Product) Use this for Causal Masking!
-
operator +(
dynamic o) → Tensor -
operator -(
dynamic o) → Tensor -
operator /(
dynamic o) → Tensor -
operator ==(
Object other) → bool -
The equality operator.
inherited
-
operator unary-(
) → Tensor
Static Methods
-
aft(
Tensor q, Tensor k, Tensor v, Tensor wb, bool masked) → Tensor -
aftCross(
Tensor q, Tensor k, Tensor v, Tensor wb) → Tensor -
concat(
List< Tensor> tensors) → Tensor -
embeddings(
List< int> idx, Tensor wte, Tensor wpe) → Tensor -
l2Normalize(
Tensor input, List< Tensor> tracker, {double eps = 1e-12}) → Tensor -
layerNorm(
Tensor x, Tensor gamma, Tensor beta, double eps) → Tensor -
random(
List< int> shape, {double scale = 0.005}) → Tensor -
zeros(
List< int> shape) → Tensor