destack:tensor

import destack:tensor

Structs

- AxisSet — Axis set used by reductions and broadcasts.
- ConvolutionAxes — Axis mapping for tensor convolution.
- ConvolutionGroups — Grouping parameters for tensor convolution.
- ConvolutionWindow — Window parameters for tensor convolution.
- Coordinate — Tensor coordinate with Rank indices.
- DotAxes — Axis mapping for tensor dot.
- FactorizationOptions — Options for matrix factorizations.
- GatherAxes — Axis mapping for tensor gather.
- NamedAxis — Named axis metadata used by frontends and debugging tools.
- NormalizationOptions — Options for normalization operations.
- Padding — Padding for a tensor with Rank axes.
- Permutation — Axis permutation with Rank entries.
- PoolOptions — Window options for pooling.
- PoolWindow — Window parameters for pooling.
- ScatterAxes — Axis mapping for tensor scatter.
- Shape — Tensor shape with Rank axes.
- SliceWindow — Slice window for a tensor with Rank axes.
- SoftmaxOptions — Options for softmax-like operations.
- TensorConvertOptions — Output options for tensor conversion.
- TensorConvolutionOptions — Output options for tensor convolution.
- TensorCreateOptions — Allocation options for a new tensor.
- TensorExecutionOptions — Execution options for tensor work.
- TensorOutputOptions — Output options for tensor operations.
- TensorReduceOptions — Output options for tensor reductions.
- TensorScatterOptions — Output options for tensor scatter.

Enums

- ConvertMode — Conversion mode for tensor element conversion.
- FactorizationMode — Matrix factorization mode.
- MemoryOrder — Memory order for dense tensor storage.
- Reduction — Reduction used when partial tensor shards are combined.
- ScatterMode — Update mode for tensor scatter.

Newtypes

- Axis — Axis index inside a tensor rank.
- Dense — Dense contiguous tensor format.
- Extent — Runtime size along one tensor axis.
- Partial — Store partial results across one mesh axis.
- Replicate — Replicate one tensor axis across the mesh.
- Shard — Split one tensor axis across one mesh axis.
- Sharding — Tensor storage partitioned or replicated across a compute mesh.
- Stride — Tensor element stride.
- Strided — Explicit strided tensor view format.
- Tensor — Owning tensor value.
- TensorView — Borrowed view over tensor storage.
- Unsharded — Tensor storage on one device.

Newtype interfaces

- Placement — Tensor placement descriptor.
- ShardingAxis — Per-axis placement descriptor used by a sharded tensor.
- TensorFormat — Format implemented by owning tensor storage.
- TensorViewFormat — Format implemented by borrowed tensor views.

Type aliases

- ColumnMajorFormat — Column-major dense tensor format.
- RowMajorFormat — Row-major dense tensor format.

Functions

- absolute — Return elementwise absolute value.
- all — Return whether all selected boolean values are true.
- any — Return whether any selected boolean value is true.
- argmax — Return indices of maximum values along axis.
- argmin — Return indices of minimum values along axis.
- avgPool — Return average-pooled values.
- bernoulli — Return values sampled from Bernoulli trials.
- bitAnd — Return bitwise AND along the axes in axisSet.
- bitOr — Return bitwise OR along the axes in axisSet.
- bitXor — Return bitwise XOR along the axes in axisSet.
- broadcast — Broadcast tensor to shape.
- cast — Cast tensor elements to U by bit reinterpretation.
- cholesky — Return the Cholesky factorization.
- concat — Concatenate tensors along axis.
- convert — Convert tensor elements to U.
- convolution — Convolve input with kernel.
- cos — Return elementwise cosine.
- det — Return the determinant of a matrix.
- diagonal — Return the diagonal of a matrix.
- dot — Multiply and reduce two tensors over configured axes.
- equal — Return elementwise equality.
- exp — Return elementwise exponential.
- expandDims — Insert a length-one axis.
- fill — Fill view with value.
- gather — Gather slices from operand.
- gelu — Return Gaussian error linear activation.
- greater — Return elementwise greater-than comparison.
- greaterEqual — Return elementwise greater-than-or-equal comparison.
- inner — Return the inner product of two vectors.
- inverse — Return the inverse of a matrix.
- layerNorm — Return layer-normalized values.
- less — Return elementwise less-than comparison.
- lessEqual — Return elementwise less-than-or-equal comparison.
- load — Load one tensor element from view.
- log — Return elementwise natural logarithm.
- logSoftmax — Return log-softmax over one axis.
- matmul — Matrix multiply rank-two tensors.
- max — Return maximum values of tensor along the axes in axisSet.
- maxPool — Return max-pooled values.
- mean — Return the mean of tensor along the axes in axisSet.
- min — Return minimum values of tensor along the axes in axisSet.
- negative — Return elementwise negation.
- norm — Return vector or matrix norm along axes.
- normal — Return values sampled from the standard normal distribution.
- notEqual — Return elementwise inequality.
- outer — Return the outer product of two vectors.
- pad — Pad tensor.
- product — Multiply elements of tensor along the axes in axisSet.
- qr — Return the QR factorization.
- reformat — Copy tensor data into a new format.
- relu — Return rectified linear activation.
- reshape — Reshape tensor to shape.
- scatter — Scatter updates into operand.
- select — Select tensor elements from thenValue or elseValue.
- sigmoid — Return logistic sigmoid activation.
- sin — Return elementwise sine.
- slice — Slice tensor by window.
- softmax — Return softmax over one axis.
- solve — Solve a linear system.
- splat — Build a tensor by repeating value.
- sqrt — Return elementwise square root.
- squeeze — Remove length-one axes.
- stack — Stack tensors along a new axis.
- store — Store one tensor element into view.
- sum — Sum tensor along the axes in axisSet.
- svd — Return the singular value decomposition.
- tan — Return elementwise tangent.
- trace — Return the trace of a matrix.
- transpose — Transpose tensor by permutation.
- uniform — Return values sampled uniformly from 0, 1).
- [uniformRange — Return values sampled uniformly from [low, high).