utils/tensor

Helper module for Tensor processing.

These functions and classes are only used internally, meaning an end-user shouldn’t need to access anything here.

• utils/tensor
  • static
    • .Tensor
      • new Tensor(...args)
      • .dims : Array.<number>
      • .type : DataType
      • .data : DataArray
      • .size : number
      • .location : string
      • .Symbol.iterator() ⇒ Iterator
      • ._getitem(index) ⇒ Tensor
      • .indexOf(item) ⇒ number
      • ._subarray(index, iterSize, iterDims) ⇒ Tensor
      • .item() ⇒ number | bigint
      • .tolist() ⇒ Array
      • .sigmoid() ⇒ Tensor
      • .sigmoid_() ⇒ Tensor
      • .map(callback) ⇒ Tensor
      • .map_(callback) ⇒ Tensor
      • .mul(val) ⇒ Tensor
      • .mul_(val) ⇒ Tensor
      • .div(val) ⇒ Tensor
      • .div_(val) ⇒ Tensor
      • .add(val) ⇒ Tensor
      • .add_(val) ⇒ Tensor
      • .sub(val) ⇒ Tensor
      • .sub_(val) ⇒ Tensor
      • .clone() ⇒ Tensor
      • .slice(...slices) ⇒ Tensor
      • .permute(...dims) ⇒ Tensor
      • .sum([dim], keepdim) ⇒
      • .norm([p], [dim], [keepdim]) ⇒ Tensor
      • .normalize_([p], [dim]) ⇒ Tensor
      • .normalize([p], [dim]) ⇒ Tensor
      • .stride() ⇒ Array.<number>
      • .squeeze([dim]) ⇒ Tensor
      • .squeeze_()
      • .unsqueeze(dim) ⇒ Tensor
      • .unsqueeze_()
      • .flatten_()
      • .flatten(start_dim, end_dim) ⇒ Tensor
      • .view(...dims) ⇒ Tensor
      • .clamp_()
      • .clamp(min, max) ⇒ Tensor
      • .round_()
      • .round() ⇒ Tensor
      • .to(type) ⇒ Tensor
    • .permute(tensor, axes) ⇒ Tensor
    • .interpolate(input, size, mode, align_corners) ⇒ Tensor
    • .interpolate_4d(input, options) ⇒ Promise.<Tensor>
    • .matmul(a, b) ⇒ Promise.<Tensor>
    • .rfft(x, a) ⇒ Promise.<Tensor>
    • .topk(x, k) ⇒ *
    • .slice(data:, starts:, ends:, axes:, [steps]) ⇒ Promise.<Tensor>
    • .mean_pooling(last_hidden_state, attention_mask) ⇒ Tensor
    • .layer_norm(input, normalized_shape, options) ⇒ Tensor
    • .cat(tensors, dim) ⇒ Tensor
    • .stack(tensors, dim) ⇒ Tensor
    • .std_mean(input, dim, correction, keepdim) ⇒ Array.<Tensor>
    • .mean(input, dim, keepdim) ⇒ Tensor
    • .full(size, fill_value) ⇒ Tensor
    • .ones(size) ⇒ Tensor
    • .ones_like(tensor) ⇒ Tensor
    • .zeros(size) ⇒ Tensor
    • .zeros_like(tensor) ⇒ Tensor
    • .rand(size) ⇒ Tensor
    • .quantize_embeddings(tensor, precision) ⇒ Tensor
  • inner
    • ~args[0] : ONNXTensor
    • ~reshape(data, dimensions) ⇒ *
      • ~reshapedArray : any
    • ~DataArray : *
    • ~NestArray : *

utils/tensor.Tensor

Kind: static class of utils/tensor

• .Tensor
  • new Tensor(...args)
  • .dims : Array.<number>
  • .type : DataType
  • .data : DataArray
  • .size : number
  • .location : string
  • .Symbol.iterator() ⇒ Iterator
  • ._getitem(index) ⇒ Tensor
  • .indexOf(item) ⇒ number
  • ._subarray(index, iterSize, iterDims) ⇒ Tensor
  • .item() ⇒ number | bigint
  • .tolist() ⇒ Array
  • .sigmoid() ⇒ Tensor
  • .sigmoid_() ⇒ Tensor
  • .map(callback) ⇒ Tensor
  • .map_(callback) ⇒ Tensor
  • .mul(val) ⇒ Tensor
  • .mul_(val) ⇒ Tensor
  • .div(val) ⇒ Tensor
  • .div_(val) ⇒ Tensor
  • .add(val) ⇒ Tensor
  • .add_(val) ⇒ Tensor
  • .sub(val) ⇒ Tensor
  • .sub_(val) ⇒ Tensor
  • .clone() ⇒ Tensor
  • .slice(...slices) ⇒ Tensor
  • .permute(...dims) ⇒ Tensor
  • .sum([dim], keepdim) ⇒
  • .norm([p], [dim], [keepdim]) ⇒ Tensor
  • .normalize_([p], [dim]) ⇒ Tensor
  • .normalize([p], [dim]) ⇒ Tensor
  • .stride() ⇒ Array.<number>
  • .squeeze([dim]) ⇒ Tensor
  • .squeeze_()
  • .unsqueeze(dim) ⇒ Tensor
  • .unsqueeze_()
  • .flatten_()
  • .flatten(start_dim, end_dim) ⇒ Tensor
  • .view(...dims) ⇒ Tensor
  • .clamp_()
  • .clamp(min, max) ⇒ Tensor
  • .round_()
  • .round() ⇒ Tensor
  • .to(type) ⇒ Tensor

new Tensor(...args)

Create a new Tensor or copy an existing Tensor.

tensor.dims : <code> Array. < number > </code>

Dimensions of the tensor.

Kind: instance property of Tensor

tensor.type : <code> DataType </code>

Type of the tensor.

Kind: instance property of Tensor

tensor.data : <code> DataArray </code>

The data stored in the tensor.

Kind: instance property of Tensor

tensor.size : <code> number </code>

The number of elements in the tensor.

Kind: instance property of Tensor

tensor.location : <code> string </code>

The location of the tensor data.

Kind: instance property of Tensor

tensor.Symbol.iterator() ⇒ <code> Iterator </code>

Returns an iterator object for iterating over the tensor data in row-major order. If the tensor has more than one dimension, the iterator will yield subarrays.

Kind: instance method of Tensor
Returns: Iterator - An iterator object for iterating over the tensor data in row-major order.

tensor._getitem(index) ⇒ <code> Tensor </code>

Index into a Tensor object.

Kind: instance method of Tensor
Returns: Tensor - The data at the specified index.

tensor.indexOf(item) ⇒ <code> number </code>

Kind: instance method of Tensor
Returns: number - The index of the first occurrence of item in the tensor data.

tensor._subarray(index, iterSize, iterDims) ⇒ <code> Tensor </code>

Kind: instance method of Tensor

tensor.item() ⇒ <code> number </code> | <code> bigint </code>

Returns the value of this tensor as a standard JavaScript Number. This only works for tensors with one element. For other cases, see Tensor.tolist().

Kind: instance method of Tensor
Returns: number | bigint - The value of this tensor as a standard JavaScript Number.
Throws:

• Error If the tensor has more than one element.

tensor.tolist() ⇒ <code> Array </code>

Convert tensor data to a n-dimensional JS list

Kind: instance method of Tensor

tensor.sigmoid() ⇒ <code> Tensor </code>

Return a new Tensor with the sigmoid function applied to each element.

Kind: instance method of Tensor
Returns: Tensor - The tensor with the sigmoid function applied.

tensor.sigmoid_() ⇒ <code> Tensor </code>

Applies the sigmoid function to the tensor in place.

Kind: instance method of Tensor
Returns: Tensor - Returns this.

tensor.map(callback) ⇒ <code> Tensor </code>

Return a new Tensor with a callback function applied to each element.

Kind: instance method of Tensor
Returns: Tensor - A new Tensor with the callback function applied to each element.

tensor.map_(callback) ⇒ <code> Tensor </code>

Apply a callback function to each element of the tensor in place.

Kind: instance method of Tensor
Returns: Tensor - Returns this.

tensor.mul(val) ⇒ <code> Tensor </code>

Return a new Tensor with every element multiplied by a constant.

Kind: instance method of Tensor
Returns: Tensor - The new tensor.

tensor.mul_(val) ⇒ <code> Tensor </code>

Multiply the tensor by a constant in place.

Kind: instance method of Tensor
Returns: Tensor - Returns this.

tensor.div(val) ⇒ <code> Tensor </code>

Return a new Tensor with every element divided by a constant.

Kind: instance method of Tensor
Returns: Tensor - The new tensor.

tensor.div_(val) ⇒ <code> Tensor </code>

Divide the tensor by a constant in place.

Kind: instance method of Tensor
Returns: Tensor - Returns this.

tensor.add(val) ⇒ <code> Tensor </code>

Return a new Tensor with every element added by a constant.

Kind: instance method of Tensor
Returns: Tensor - The new tensor.

tensor.add_(val) ⇒ <code> Tensor </code>

Add the tensor by a constant in place.

Kind: instance method of Tensor
Returns: Tensor - Returns this.

tensor.sub(val) ⇒ <code> Tensor </code>

Return a new Tensor with every element subtracted by a constant.

Kind: instance method of Tensor
Returns: Tensor - The new tensor.

tensor.sub_(val) ⇒ <code> Tensor </code>

Subtract the tensor by a constant in place.

Kind: instance method of Tensor
Returns: Tensor - Returns this.

tensor.clone() ⇒ <code> Tensor </code>

Creates a deep copy of the current Tensor.

Kind: instance method of Tensor
Returns: Tensor - A new Tensor with the same type, data, and dimensions as the original.

tensor.slice(...slices) ⇒ <code> Tensor </code>

Performs a slice operation on the Tensor along specified dimensions.

Consider a Tensor that has a dimension of [4, 7]:

[ 1,  2,  3,  4,  5,  6,  7]
[ 8,  9, 10, 11, 12, 13, 14]
[15, 16, 17, 18, 19, 20, 21]
[22, 23, 24, 25, 26, 27, 28]

We can slice against the two dims of row and column, for instance in this case we can start at the second element, and return to the second last, like this:

tensor.slice([1, -1], [1, -1]);

which would return:

[  9, 10, 11, 12, 13 ]
[ 16, 17, 18, 19, 20 ]

Kind: instance method of Tensor
Returns: Tensor - A new Tensor containing the selected elements.
Throws:

• Error If the slice input is invalid.

tensor.permute(...dims) ⇒ <code> Tensor </code>

Return a permuted version of this Tensor, according to the provided dimensions.

Kind: instance method of Tensor
Returns: Tensor - The permuted tensor.

tensor.sum([dim], keepdim) ⇒

Returns the sum of each row of the input tensor in the given dimension dim.

Kind: instance method of Tensor
Returns: The summed tensor

tensor.norm([p], [dim], [keepdim]) ⇒ <code> Tensor </code>

Returns the matrix norm or vector norm of a given tensor.

Kind: instance method of Tensor
Returns: Tensor - The norm of the tensor.

tensor.normalize_([p], [dim]) ⇒ <code> Tensor </code>

Performs L_p normalization of inputs over specified dimension. Operates in place.

Kind: instance method of Tensor
Returns: Tensor - this for operation chaining.

tensor.normalize([p], [dim]) ⇒ <code> Tensor </code>

Performs L_p normalization of inputs over specified dimension.

Kind: instance method of Tensor
Returns: Tensor - The normalized tensor.

tensor.stride() ⇒ <code> Array. < number > </code>

Compute and return the stride of this tensor. Stride is the jump necessary to go from one element to the next one in the specified dimension dim.

Kind: instance method of Tensor
Returns: Array.<number> - The stride of this tensor.

tensor.squeeze([dim]) ⇒ <code> Tensor </code>

Returns a tensor with all specified dimensions of input of size 1 removed.

NOTE: The returned tensor shares the storage with the input tensor, so changing the contents of one will change the contents of the other. If you would like a copy, use tensor.clone() before squeezing.

Kind: instance method of Tensor
Returns: Tensor - The squeezed tensor

tensor.squeeze_()

In-place version of @see Tensor.squeeze

Kind: instance method of Tensor

tensor.unsqueeze(dim) ⇒ <code> Tensor </code>

Returns a new tensor with a dimension of size one inserted at the specified position.

NOTE: The returned tensor shares the same underlying data with this tensor.

Kind: instance method of Tensor
Returns: Tensor - The unsqueezed tensor

tensor.unsqueeze_()

In-place version of @see Tensor.unsqueeze

Kind: instance method of Tensor

tensor.flatten_()

In-place version of @see Tensor.flatten

Kind: instance method of Tensor

tensor.flatten(start_dim, end_dim) ⇒ <code> Tensor </code>

Flattens input by reshaping it into a one-dimensional tensor. If start_dim or end_dim are passed, only dimensions starting with start_dim and ending with end_dim are flattened. The order of elements in input is unchanged.

Kind: instance method of Tensor
Returns: Tensor - The flattened tensor.

tensor.view(...dims) ⇒ <code> Tensor </code>

Returns a new tensor with the same data as the self tensor but of a different shape.

Kind: instance method of Tensor
Returns: Tensor - The tensor with the same data but different shape

tensor.clamp_()

In-place version of @see Tensor.clamp

Kind: instance method of Tensor

tensor.clamp(min, max) ⇒ <code> Tensor </code>

Clamps all elements in input into the range [ min, max ]

Kind: instance method of Tensor
Returns: Tensor - the output tensor.

tensor.round_()

In-place version of @see Tensor.round

Kind: instance method of Tensor

tensor.round() ⇒ <code> Tensor </code>

Rounds elements of input to the nearest integer.

Kind: instance method of Tensor
Returns: Tensor - the output tensor.

tensor.to(type) ⇒ <code> Tensor </code>

Performs Tensor dtype conversion.

Kind: instance method of Tensor
Returns: Tensor - The converted tensor.

utils/tensor.permute(tensor, axes) ⇒ <code> Tensor </code>

Permutes a tensor according to the provided axes.

Kind: static method of utils/tensor
Returns: Tensor - The permuted tensor.

utils/tensor.interpolate(input, size, mode, align_corners) ⇒ <code> Tensor </code>

Interpolates an Tensor to the given size.

Kind: static method of utils/tensor
Returns: Tensor - The interpolated tensor.

utils/tensor.interpolate_4d(input, options) ⇒ <code> Promise. < Tensor > </code>

Down/up samples the input. Inspired by https://pytorch.org/docs/stable/generated/torch.nn.functional.interpolate.html.

Kind: static method of utils/tensor
Returns: Promise.<Tensor> - The interpolated tensor.

utils/tensor.matmul(a, b) ⇒ <code> Promise. < Tensor > </code>

Matrix product of two tensors. Inspired by https://pytorch.org/docs/stable/generated/torch.matmul.html

Kind: static method of utils/tensor
Returns: Promise.<Tensor> - The matrix product of the two tensors.

utils/tensor.rfft(x, a) ⇒ <code> Promise. < Tensor > </code>

Computes the one dimensional Fourier transform of real-valued input. Inspired by https://pytorch.org/docs/stable/generated/torch.fft.rfft.html

Kind: static method of utils/tensor
Returns: Promise.<Tensor> - the output tensor.

utils/tensor.topk(x, k) ⇒ <code> * </code>

Returns the k largest elements of the given input tensor. Inspired by https://pytorch.org/docs/stable/generated/torch.topk.html

Kind: static method of utils/tensor
Returns: * - the output tuple of (Tensor, LongTensor) of top-k elements and their indices.

utils/tensor.slice(data:, starts:, ends:, axes:, [steps]) ⇒ <code> Promise. < Tensor > </code>

Slice a multidimensional float32 tensor.

Kind: static method of utils/tensor
Returns: Promise.<Tensor> - Sliced data tensor.

utils/tensor.mean_pooling(last_hidden_state, attention_mask) ⇒ <code> Tensor </code>

Perform mean pooling of the last hidden state followed by a normalization step.

Kind: static method of utils/tensor
Returns: Tensor - Returns a new Tensor of shape [batchSize, embedDim].

utils/tensor.layer_norm(input, normalized_shape, options) ⇒ <code> Tensor </code>

Apply Layer Normalization for last certain number of dimensions.

Kind: static method of utils/tensor
Returns: Tensor - The normalized tensor.

utils/tensor.cat(tensors, dim) ⇒ <code> Tensor </code>

Concatenates an array of tensors along a specified dimension.

Kind: static method of utils/tensor
Returns: Tensor - The concatenated tensor.

utils/tensor.stack(tensors, dim) ⇒ <code> Tensor </code>

Stack an array of tensors along a specified dimension.

Kind: static method of utils/tensor
Returns: Tensor - The stacked tensor.

utils/tensor.std_mean(input, dim, correction, keepdim) ⇒ <code> Array. < Tensor > </code>

Calculates the standard deviation and mean over the dimensions specified by dim. dim can be a single dimension or null to reduce over all dimensions.

Kind: static method of utils/tensor
Returns: Array.<Tensor> - A tuple of (std, mean) tensors.

utils/tensor.mean(input, dim, keepdim) ⇒ <code> Tensor </code>

Returns the mean value of each row of the input tensor in the given dimension dim.

Kind: static method of utils/tensor
Returns: Tensor - A new tensor with means taken along the specified dimension.

utils/tensor.full(size, fill_value) ⇒ <code> Tensor </code>

Creates a tensor of size size filled with fill_value. The tensor’s dtype is inferred from fill_value.

Kind: static method of utils/tensor
Returns: Tensor - The filled tensor.

utils/tensor.ones(size) ⇒ <code> Tensor </code>

Returns a tensor filled with the scalar value 1, with the shape defined by the variable argument size.

Kind: static method of utils/tensor
Returns: Tensor - The ones tensor.

utils/tensor.ones_like(tensor) ⇒ <code> Tensor </code>

Returns a tensor filled with the scalar value 1, with the same size as input.

Kind: static method of utils/tensor
Returns: Tensor - The ones tensor.

utils/tensor.zeros(size) ⇒ <code> Tensor </code>

Returns a tensor filled with the scalar value 0, with the shape defined by the variable argument size.

Kind: static method of utils/tensor
Returns: Tensor - The zeros tensor.

utils/tensor.zeros_like(tensor) ⇒ <code> Tensor </code>

Returns a tensor filled with the scalar value 0, with the same size as input.

Kind: static method of utils/tensor
Returns: Tensor - The zeros tensor.

utils/tensor.rand(size) ⇒ <code> Tensor </code>

Returns a tensor filled with random numbers from a uniform distribution on the interval [0, 1)

Kind: static method of utils/tensor
Returns: Tensor - The random tensor.

utils/tensor.quantize_embeddings(tensor, precision) ⇒ <code> Tensor </code>

Quantizes the embeddings tensor to binary or unsigned binary precision.

Kind: static method of utils/tensor
Returns: Tensor - The quantized tensor.

utils/tensor~args[0] : <code> ONNXTensor </code>

Kind: inner property of utils/tensor

utils/tensor~reshape(data, dimensions) ⇒ <code> * </code>

Reshapes a 1-dimensional array into an n-dimensional array, according to the provided dimensions.

Kind: inner method of utils/tensor
Returns: * - The reshaped array.

Example

reshape([10                    ], [1      ]); // Type: number[]      Value: [10]
  reshape([1, 2, 3, 4            ], [2, 2   ]); // Type: number[][]    Value: [[1, 2], [3, 4]]
  reshape([1, 2, 3, 4, 5, 6, 7, 8], [2, 2, 2]); // Type: number[][][]  Value: [[[1, 2], [3, 4]], [[5, 6], [7, 8]]]
  reshape([1, 2, 3, 4, 5, 6, 7, 8], [4, 2   ]); // Type: number[][]    Value: [[1, 2], [3, 4], [5, 6], [7, 8]]

reshape~reshapedArray : <code> any </code>

Kind: inner property of reshape

utils/tensor~DataArray : <code> * </code>

Kind: inner typedef of utils/tensor

utils/tensor~NestArray : <code> * </code>

This creates a nested array of a given type and depth (see examples).

Kind: inner typedef of utils/tensor
Example

NestArray<string, 1>; // string[]

Example

NestArray<number, 2>; // number[][]

Example

NestArray<string, 3>; // string[][][] etc.