Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance. Project price only 1 $
You can buy this project and download/modify it how often you want.
/*
* Copyright 2016 The BigDL Authors.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.intel.analytics.bigdl.tensor
import java.io.Serializable
import breeze.linalg.{DenseMatrix => BrzDenseMatrix, DenseVector => BrzDenseVector}
import com.intel.analytics.bigdl.mkl.MKL
import com.intel.analytics.bigdl.nn.abstractnn.Activity
import com.intel.analytics.bigdl.tensor.TensorNumericMath.TensorNumeric
import com.intel.analytics.bigdl.utils.{File, Table}
import org.apache.spark.mllib.linalg.{DenseMatrix, DenseVector, Matrix, Vector}
import scala.collection.mutable
import scala.collection.mutable.ArrayBuffer
import scala.reflect.ClassTag
/**
* It is the class for handling numeric data.
*
* @tparam T should be Double or Float
*/
trait Tensor[T] extends Serializable with TensorMath[T] with Activity {
/**
* @return whether this tensor is an empty tensor. Note that nDimension == 0 is not
* sufficient to determine a tensor is empty, because a scalar tensor's nDimension
* is also 0.
*/
def isEmpty: Boolean
/**
* @return whether this tensor is a scalar
*/
def isScalar: Boolean
/**
* Dimension number of the tensor. For empty tensor, its dimension number is 0
*
* @return dimension number
*/
def nDimension(): Int
/**
* A shortcut of nDimension()
*
* @see nDimension()
*/
def dim(): Int
/**
* Size of tensor. Return an array of which each value represents the size on the
* dimension(i + 1), i is the index of the corresponding value.
* It will generate a new array each time method is invoked.
*
* @return size array
*/
def size(): Array[Int]
/**
* size of the tensor on the given dimension
*
* @param dim dimension, count from 1
* @return size
*/
def size(dim: Int): Int
/**
* Jumps between elements on the each dimension in the storage.
* It will generate a new array each time method is invoked.
*
* @return strides array
*/
def stride(): Array[Int]
/**
* Jumps between elements on the given dimension in the storage.
*
* @param dim dimension, count from 1
* @return jump
*/
def stride(dim: Int): Int
/**
* Fill with a given value. It will change the value of the current tensor and return itself
*
* @param v value to fill the tensor
* @return current tensor
*/
def fill(v: T): Tensor[T]
/**
* Fill with a given value. It will change the value of the current tensor and return itself
*
* Note the value should be an instance of T
*
* @param v value to fill the tensor
* @return current tensor
*/
def forceFill(v: Any): Tensor[T]
/**
* Fill with zero. It will change the value of the current tensor and return itself
*
* @return current tensor
*/
def zero(): Tensor[T]
/**
* Fill with random value(normal gaussian distribution).
* It will change the value of the current tensor and return itself
*
* @return current tensor
*/
def randn(): Tensor[T]
/**
* Fill with random value(normal gaussian distribution with the specified mean
* and stdv).
* It will change the value of the current tensor and return itself
*
* @return current tensor
*/
def randn(mean: Double, stdv: Double): Tensor[T]
/**
* Fill with random value(uniform distribution).
* It will change the value of the current tensor and return itself
*
* @return current tensor
*/
def rand(): Tensor[T]
/**
* Fill with random value(uniform distribution between [lowerBound, upperBound])
* It will change the value of the current tensor and return itself
*
* @return current tensor
*/
def rand(lowerBound: Double, upperBound: Double): Tensor[T]
/**
* Fill with random value(bernoulli distribution).
* It will change the value of the current tensor and return itself
*
* @return current tensor
*/
def bernoulli(p: Double): Tensor[T]
/** *
* Create a new tensor which exchanges the given dimensions of the current tensor
*
* @param dim1 dimension to be exchanged, count from one
* @param dim2 dimension to be exchanged, count from one
* @return new tensor
*/
def transpose(dim1: Int, dim2: Int): Tensor[T]
/**
* Shortcut of transpose(1, 2) for 2D tensor
*
* @see transpose()
*/
def t(): Tensor[T]
/**
* Query tensor on a given index. Tensor should not be empty
*
* @param index count from 1
* @return
*/
def apply(index: Int): Tensor[T]
/**
* Query the value on a given index. Tensor should not be empty
*
* @param indexes the indexes length should be same as the tensor dimension length and each
* value count from 1
* @return the value on the given index
*/
def apply(indexes: Array[Int]): T
/**
* @return the value of a scalar. Requires the tensor to be a scalar.
*/
def value(): T
/**
* Query the value on a given position. The number of parameters
* should be equal to the dimension number of the tensor.
* Tensor should not be empty.
*
* @param d1,( d2, d3, d4, d5) the given position
* @return the value on a given position
*/
def valueAt(d1: Int): T
def valueAt(d1: Int, d2: Int): T
def valueAt(d1: Int, d2: Int, d3: Int): T
def valueAt(d1: Int, d2: Int, d3: Int, d4: Int): T
def valueAt(d1: Int, d2: Int, d3: Int, d4: Int, d5: Int): T
/**
* Subset the tensor by apply the elements of the given table to the corresponding dimension
* of the tensor. The elements of the given table can be an Int or another Table.
* An Int means select on current dimension; A table means narrow on current dimension,
* the table should have two elements, of which the first is the start index and
* the second is the end index. An empty table is equal to Table(1, size_of_current_dimension)
* If the table length is less than the tensor dimension, each missing dimension is token up by
* an empty table
*
* @see select
* @see narrow
* @param t The table length should be less than or equal to the tensor dimensions
* @return
*/
def apply(t: Table): Tensor[T]
/**
* For tensor(i) = value. If tensor(i) is another tensor, it will fill the selected subset by
* the given value
*
* @param index index
* @param value value to write
*/
def update(index: Int, value: T): Unit
/**
* Copy the give tensor value to the select subset of the current tensor by the given index.
* The subset should have the same size of the given tensor
*
* @param index index
* @param src tensor to write
*/
def update(index: Int, src: Tensor[T]): Unit
/**
* Write the value to the positions indexed by the given index array
*
* @param indexes index array. It should has same length with the tensor dimension
* @param value value to write
*/
def update(indexes: Array[Int], value: T): Unit
/**
* Set value for a scalar tensor
* @param value the written value
* @return
*/
def setValue(value: T): this.type
/**
* Write the value on a given position. The number of parameters
* should be equal to the dimension number of the tensor.
*
* @param d1,( d2, d3, d4, d5) the given position
* @param value the written value
* @return
*/
def setValue(d1: Int, value: T): this.type
def setValue(d1: Int, d2: Int, value: T): this.type
def setValue(d1: Int, d2: Int, d3: Int, value: T): this.type
def setValue(d1: Int, d2: Int, d3: Int, d4: Int, value: T): this.type
def setValue(d1: Int, d2: Int, d3: Int, d4: Int, d5: Int, value: T): this.type
/**
* Fill the select subset of the current tensor with the given value.
* The element of the given table can be an Int or another Table. An Int means select on current
* dimension; A table means narrow on the current dimension, the table should has two elements,
* of which the first is the start index and the second is the end index. An empty table is equal
* to Table(1, size_of_current_dimension) If the table length is less than the tensor dimension,
* each missing dimension is applied by an empty table
*
* @param t subset table
* @param value value to write
*/
def update(t: Table, value: T): Unit
/**
* Copy the given tensor values to the selected subset of the current tensor
* Each element of the given table can be an Int or another Table. An Int means select on current
* dimension; A table means narrow on current dimension, the table should has two elements,
* of which the first is start index and the second is the end index. An empty table is equal
* to Table(1, size_of_current_dimension). If the table's length is smaller than the tensor's
* dimension, the missing dimension is applied by an empty table.
*
* @param t subset table
* @param src tensor to copy
*/
def update(t: Table, src: Tensor[T]): Unit
/**
* Update the value meeting the filter criteria with the give value
*
* @param filter filter
* @param value value to update
*/
def update(filter: T => Boolean, value: T): Unit
/**
* Check if the tensor is contiguous on the storage
*
* @return true if it's contiguous
*/
def isContiguous(): Boolean
/**
* Get a contiguous tensor from current tensor
*
* @return the current tensor if it's contiguous; or a new contiguous tensor with separated
* storage
*/
def contiguous(): Tensor[T]
/**
* Check if the size is same with the give tensor
*
* @param other tensor to be compared
* @return true if they have same size
*/
def isSameSizeAs(other: Tensor[_]): Boolean
/**
* Get a new tensor with same value and different storage
*
* @return new tensor
*/
override def clone(): Tensor[T] = {
this
}
/**
* Get a new tensor with same storage.
*
* @return new tensor
*/
def shallowClone(): Tensor[T] = {
this
}
/**
* return a new empty tensor of the same type
*
* @return new tensor
*/
def emptyInstance(): Tensor[T]
/**
* Resize the current tensor to the same size of the given tensor. It will still use the same
* storage if the storage
* is sufficient for the new size
*
* @param src target tensor
* @return current tensor
*/
def resizeAs(src: Tensor[_]): Tensor[T]
/**
* Cast the currenct tensor to a tensor with tensor numeric type D
* and set cast value to `castTensor`
*
* @param castTensor the cast value set to this tensor
* @tparam D new numeric type
* @return return castTensort
*/
def cast[D: ClassTag](castTensor: Tensor[D])(implicit ev: TensorNumeric[D]): Tensor[D]
/**
* Resize the current tensor to the give shape
*
* @param sizes Array describe the size
* @param strides Array describe the jumps
* @return
*/
def resize(sizes: Array[Int], strides: Array[Int] = null): Tensor[T]
def resize(size1: Int): Tensor[T]
def resize(size1: Int, size2: Int): Tensor[T]
def resize(size1: Int, size2: Int, size3: Int): Tensor[T]
def resize(size1: Int, size2: Int, size3: Int, size4: Int): Tensor[T]
def resize(size1: Int, size2: Int, size3: Int, size4: Int, size5: Int): Tensor[T]
def resize(sizes: Array[Int], nElement: Int): Tensor[T] = {
throw new UnsupportedOperationException("resize with nElement for sparse tensor only")
}
// def repeatTensor(result: Tensor, tensor: Tensor, size: Int*)
/**
* Element number
*
* @return element number
*/
def nElement(): Int
/**
* Remove the dim-th dimension and return the subset part. For instance
* tensor =
* 1 2 3
* 4 5 6
* tensor.select(1, 1) is [1 2 3]
* tensor.select(1, 2) is [4 5 6]
* tensor.select(2, 3) is [3 6]
*
* @param dim
* @param index
* @return
*/
def select(dim: Int, index: Int): Tensor[T]
/**
* Get the storage
*
* @return storage
*/
def storage(): Storage[T]
/**
* tensor offset on the storage
*
* @return storage offset, count from 1
*/
def storageOffset(): Int
/**
* The Tensor is now going to "view" the same storage as the given tensor. As the result,
* any modification in the elements of the Tensor will have an impact on the elements of the
* given tensor, and vice-versa. This is an efficient method, as there is no memory copy!
*
* @param other the given tensor
* @return current tensor
*/
def set(other: Tensor[T]): Tensor[T]
/**
* The Tensor is now going to "view" the given storage, starting at position storageOffset (>=1)
* with the given dimension sizes and the optional given strides. As the result, any
* modification in the elements of the Storage will have an impact on the elements of the Tensor,
* and vice-versa. This is an efficient method, as there is no memory copy!
*
* If only storage is provided, the whole storage will be viewed as a 1D Tensor.
*
* @param storage
* @param storageOffset
* @param sizes
* @param strides
* @return current tensor
*/
def set(storage: Storage[T], storageOffset: Int = 1, sizes: Array[Int] = null,
strides: Array[Int] = null): Tensor[T]
/**
* Shrunk the size of the storage to 0, and also the tensor size
*
* @return
*/
def set(): Tensor[T]
/**
* Get a subset of the tensor on dim-th dimension. The offset is given by index, and length is
* given by size. The important difference with select is that it will not reduce the dimension
* number. For Instance
* tensor =
* 1 2 3
* 4 5 6
* tensor.narrow(1, 1, 1) is [1 2 3]
* tensor.narrow(2, 2, 2) is
* 2 3
* 5 6
*
* @param dim
* @param index
* @param size
* @return
*/
def narrow(dim: Int, index: Int, size: Int): Tensor[T]
/**
* Copy the value of the given tensor to the current. They should have same size. It will use
* the old storage
*
* @param other source tensor
* @return current tensor
*/
def copy(other: Tensor[T]): Tensor[T]
/**
* Apply a function to each element of the tensor `t`
* and set each value to self
*
* @param t tensor to be modified
* @param func applied function
* @return current tensor
*/
def applyFun[A: ClassTag](
t: Tensor[A],
func: (A) => T): Tensor[T]
/**
* Apply a function to each element of the tensor and modified it value if it return a double
*
* @param func applied function
* @return current tensor
*/
def apply1(func: T => T): Tensor[T]
/**
* Zip values of two other tensors with applying the function `func` on
* each two values element-wisely and assign the result value to the
* current tensor
*
* The two given tensors should has the same size of the current tensor
*
* @param t1 tensor 1
* @param t2 tensor 2
* @param func zip with the function
* @tparam A numeric type of tensor 1
* @tparam B numeric type of tensor 2
*
* @return self
*/
def zipWith[A: ClassTag, B: ClassTag](
t1: Tensor[A],
t2: Tensor[B],
func: (A, B) => T): Tensor[T]
/**
* Map value of another tensor to corresponding value of current tensor and apply function on
* the two value and change the value of the current tensor
* The another tensor should has the same size of the current tensor
*
* @param other another tensor
* @param func applied function
* @return current tensor
*/
def map(other: Tensor[T], func: (T, T) => T): Tensor[T]
/**
* Removes all singleton dimensions of the tensor
*
* @return current tensor
*/
def squeeze(): Tensor[T]
/**
* Removes given dimensions of the tensor if it's singleton
*
* @return current tensor
*/
def squeeze(dim: Int): Tensor[T]
/**
* Create a new tensor that removes all singleton dimensions of the tensor
*
* @return create a new tensor
*/
def squeezeNewTensor(): Tensor[T]
/**
* Return a new tensor with specified sizes. The input tensor must be contiguous, and the
* elements number in the given sizes must be equal to the current tensor
*
* @param sizes
* @return new tensor
*/
def view(sizes: Int*): Tensor[T] = {
view(sizes.toArray)
}
def view(sizes: Array[Int]): Tensor[T]
/**
* Count the number of non-zero elements in first dimension.
* For SparseTensor only.
* @return an array number of non-zero elements in first dimension.
*/
def numNonZeroByRow(): Array[Int] = {
throw new UnsupportedOperationException("countNonZero for sparse tensor only")
}
/**
*
* Returns a tensor which contains all slices of size @param size
* in the dimension @param dim. Step between two slices is given by @param step.
*
* @param dim
* @param size
* @param step Step between two slices
* @return new tensor
*/
def unfold(dim: Int, size: Int, step: Int): Tensor[T]
/**
* Repeating a tensor allocates new memory, unless result is provided, in which case its memory
* is resized. sizes specify the number of times the tensor is repeated in each dimension.
*
* @param sizes
* @return
*/
def repeatTensor(sizes: Array[Int]): Tensor[T]
/**
* This is equivalent to this.expand(template.size())
*
* @param template the given tensor
* @return
*/
def expandAs(template: Tensor[T]): Tensor[T]
/**
* Expanding a tensor allocates new memory, tensor where singleton dimensions can be expanded
* to multiple ones by setting the stride to 0. Any dimension that has size 1 can be expanded
* to arbitrary value with new memory allocation. Attempting to expand along a dimension that
* does not have size 1 will result in an error.
*
* @param sizes the size that tensor will expend to
* @return
*/
def expand(sizes: Array[Int]): Tensor[T]
/**
* Splits current tensor along dimension dim into a result table of Tensors of size size
* (a number) or less (in the case of the last Tensor). The sizes of the non-dim dimensions
* remain unchanged. Internally, a series of narrows are performed along dimensions dim.
* Argument dim defaults to 1.
*
* @param size
* @param dim
* @return
*/
def split(size: Int, dim: Int = 1): Array[Tensor[T]]
/**
* spilt one tensor into multi tensor along the `dim` dimension
* @param dim the specific dimension
* @return
*/
def split(dim: Int) : Array[Tensor[T]]
/**
* convert the tensor to BreezeVector, the dimension of the tensor need to be 1.
*
* @return BrzDenseVector
*/
def toBreezeVector(): BrzDenseVector[T]
/**
* convert the tensor to MLlibVector, the dimension of the
* tensor need to be 1, and tensor need to be continuous.
*
* @return Vector
*/
def toMLlibVector(): Vector
/**
* convert the tensor to BreezeMatrix, the dimension of the tensor need to be 2.
*
* @return BrzDenseMatrix
*/
def toBreezeMatrix(): BrzDenseMatrix[T]
/**
* convert the tensor to MLlibMatrix, the dimension of the
* tensor need to be 2, and tensor need to be continuous.
*
* @return Matrix
*/
def toMLlibMatrix(): Matrix
/**
* return the tensor datatype( DoubleType or FloatType)
*
* @return
*/
def getType(): TensorDataType
/**
* Compare and print differences between two tensors
*
* @param other
* @param count
* @return true if there's difference, vice versa
*/
def diff(other: Tensor[T], count: Int = 1, reverse: Boolean = false): Boolean
/**
* view this.tensor and add a Singleton Dimension to `dim` dimension
*
* @param t source tensor
* @param dim the specific dimension, default is 1
* @return this
*/
def addSingletonDimension(t: Tensor[T] = this, dim: Int = 1): Tensor[T]
/**
* create a new tensor without any change of the tensor
*
* @param sizes the size of the new Tensor
* @return
*/
def reshape(sizes: Array[Int]): Tensor[T]
/**
* Save the tensor to given path
*
* @param path
* @param overWrite
* @return
*/
def save(path : String, overWrite : Boolean = false) : this.type
override def toTable: Table =
throw new IllegalArgumentException("Tensor cannot be cast to Table")
/**
* Return true because it's a Tensor implemented from [[Activity]]
*
* @return true
*/
override def isTensor: Boolean = true
/**
* Return false because it's not a Table
*
* @return false
*/
override def isTable: Boolean = false
/**
* Return tensor numeric
* @return
*/
def getTensorNumeric(): TensorNumeric[T]
/**
* Return tensor type
* @return Dense / Quant
*/
def getTensorType: TensorType
/**
* Compare with other tensor. The shape of the other tensor must be same with this tensor.
* If element wise difference is less than delta, return true.
* @param other
* @param delta
* @return
*/
def almostEqual(other: Tensor[T], delta : Double): Boolean = {
var result = true
this.map(other, (a, b) => {
val tn = getTensorNumeric()
if (tn.isGreater(tn.abs(tn.minus(a, b)), tn.fromType(delta))) {
result = false
}
a
})
return result
}
/**
* Convert 1D tensor to an array. If the tensor is not 1D, an exception will be thrown out.
* @return
*/
def toArray(): Array[T]
/**
* Element wise inequality between tensor and given value
* @param value
* @return
*/
def notEqualValue(value : Double): Boolean = {
var j = 0
while (j < this.nElement()) {
if (this.storage.apply(j + this.storageOffset() - 1) != value) {
return true
}
j += 1
}
return false
}
private[bigdl] def toQuantizedTensor: QuantizedTensor[T]
}
/**
* Numeric type of tensor.
*/
sealed trait TensorDataType extends Serializable
object BooleanType extends TensorDataType
object CharType extends TensorDataType
object StringType extends TensorDataType
object IntType extends TensorDataType
object ShortType extends TensorDataType
object LongType extends TensorDataType
object FloatType extends TensorDataType
object DoubleType extends TensorDataType
sealed trait TensorType
object DenseType extends TensorType
object SparseType extends TensorType
object QuantizedType extends TensorType
object MklDnnType extends TensorType
object Tensor {
// pre-load MKL library. If we do not do it here,
// libjmkl.so will be loaded when one of the methods of in MKL is called.
MKL.isMKLLoaded
/**
* Start index in BigDL. We count from 1.
*/
val START_INDEX = 1
/**
* Returns an empty tensor.
*
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag]()(
implicit ev: TensorNumeric[T]): Tensor[T] = new DenseTensor[T]()
/**
* Create a tensor up to 5 dimensions. The tensor size will be `d1 x d2 x d3 x d4 x d5`.
*
* @param d1,(d2, d3, d4, d5)
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag](d1: Int)(
implicit ev: TensorNumeric[T]): Tensor[T] = new DenseTensor[T](d1)
def apply[@specialized(Float, Double) T: ClassTag](d1: Int, d2: Int)(
implicit ev: TensorNumeric[T]): Tensor[T] = new DenseTensor[T](d1, d2)
def apply[@specialized(Float, Double) T: ClassTag](d1: Int, d2: Int, d3: Int)(
implicit ev: TensorNumeric[T]): Tensor[T] = new DenseTensor[T](d1, d2, d3)
def apply[@specialized(Float, Double) T: ClassTag](d1: Int, d2: Int, d3: Int, d4: Int)(
implicit ev: TensorNumeric[T]): Tensor[T] = new DenseTensor[T](d1, d2, d3, d4)
def apply[@specialized(Float, Double) T: ClassTag](d1: Int, d2: Int, d3: Int, d4: Int, d5: Int)(
implicit ev: TensorNumeric[T]): Tensor[T] = new DenseTensor[T](d1, d2, d3, d4, d5)
/**
* Create a tensor with a table
* @param xs the table contains a multi-dimensional numbers
* @return a new Tensor
*/
def apply[@specialized(Float, Double) T: ClassTag](xs : Table)(
implicit ev: TensorNumeric[T]): Tensor[T] = {
val flatTable = xs.flatten()
val content = new Array[T](flatTable.length())
for (i <- 1 to content.length) {
content(i - 1) = flatTable[Any](i) match {
case e: Boolean => ev.fromType(e)
case e: Char => ev.fromType(e)
case e: Short => ev.fromType(e)
case e: Int => ev.fromType(e)
case e: Long => ev.fromType(e)
case e: Float => ev.fromType(e)
case e: Double => ev.fromType(e)
case e: String => ev.fromType(e)
case _ => throw new IllegalArgumentException(s"Not support numeric type " +
flatTable[Any](i).getClass.getName)
}
}
val dims = new ArrayBuffer[Int]()
def getDims(xs: Table): ArrayBuffer[Int] = xs match {
case _ if xs.length() != 0 =>
dims.append(xs.length())
if (xs(1).isInstanceOf[Table]) {
getDims(xs(1))
}
dims
case otherwise => dims
}
getDims(xs)
new DenseTensor[T](
new ArrayStorage[T](content), 0, dims.toArray,
DenseTensor.size2Stride(dims.toArray), dims.length)
}
/**
* Create a tensor on given dimensions. The tensor size will be the product of dims
*
* @param dims
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag](dims: Int*)(
implicit ev: TensorNumeric[T]): Tensor[T] =
new DenseTensor[T](new ArrayStorage[T](new Array[T](dims.product)), 0, dims.toArray,
DenseTensor.size2Stride(dims.toArray), dims.length)
/**
* Create a tensor on given sizes. The tensor size will be the product of sizes
*
* @param sizes
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag](sizes: Array[Int])(
implicit ev: TensorNumeric[T]): Tensor[T] =
new DenseTensor(new ArrayStorage[T](new Array[T](sizes.product)), 0, sizes.clone(),
DenseTensor.size2Stride(sizes.clone()), sizes.length)
/**
* Returns a tensor which uses the existing Storage storage.
*
* @param storage the given storage
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag](storage: Storage[T])(
implicit ev: TensorNumeric[T]): Tensor[T] = {
require(storage.isInstanceOf[ArrayStorage[_]], "Only support array storage in this operaiton")
new DenseTensor(storage.asInstanceOf[ArrayStorage[T]])
}
/**
* Returns a tensor with the given array and shape
*
* @param data the given storage
* @param shape the given shape
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag](data: Array[T],
shape: Array[Int])(implicit ev: TensorNumeric[T]): Tensor[T] = {
if (shape.product != data.length) {
require(data.length == 1, "shape total size doesn't match data length")
// Here we create a repeat tensor
val strides = new Array[Int](shape.length)
new DenseTensor[T]().set(Storage[T](data), storageOffset = 1, sizes = shape,
strides = strides)
} else {
new DenseTensor[T]().set(Storage[T](data), storageOffset = 1, sizes = shape)
}
}
/**
* Returns a tensor which uses the existing Storage storage, starting at
* position storageOffset (>=1). The size of each dimension of the tensor
* is given by the optional Array size. If not given, the size will be computed
* as the length of storage. The jump necessary to go from one element to the
* next one in each dimension is given by the optional Array stride. If not
* given, the stride() will be computed such that the tensor is as contiguous
* as possible in memory.
*
* @param storage
* @param storageOffset
* @param size
* @param stride
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag](
storage: Storage[T],
storageOffset: Int,
size: Array[Int] = null,
stride: Array[Int] = null)(implicit ev: TensorNumeric[T]): Tensor[T] = {
new DenseTensor(storage.asInstanceOf[ArrayStorage[T]], storageOffset, size, stride)
}
/**
* create a tensor with a given tensor. The tensor will have same size
* with the given tensor.
*
* @param other the given tensor
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag](other: Tensor[T])(
implicit ev: TensorNumeric[T]): Tensor[T] = new DenseTensor(other)
/**
* create a tensor with a given breeze vector. The tensor will have the same size
* with the given breeze vector.
*
* @param vector the given breeze vector
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag](vector: BrzDenseVector[T])(
implicit ev: TensorNumeric[T]): Tensor[T] = apply(Storage(vector.data),
vector.offset + 1, Array(vector.length), Array(vector.stride))
/**
* create a tensor with a given spark Densevector. The tensor will have the same size
* with the given spark Densevector.
*
* @param vector the given spark Densevector
* @return
*/
def apply(vector: DenseVector): Tensor[Double] =
apply[Double](Storage(vector.toArray))
/**
* create a tensor with a given breeze matrix. The tensor will have the same size with
* the given breeze matrix.
*
* @param matrix the given breeze matrix
* @param ev
* @tparam T
* @return
*/
def apply[@specialized(Float, Double) T: ClassTag](matrix: BrzDenseMatrix[T])(
implicit ev: TensorNumeric[T]): Tensor[T] = apply(Storage(matrix.data),
matrix.offset + 1, Array(matrix.rows, matrix.cols),
if (matrix.isTranspose) Array(matrix.majorStride, 1) else Array(1, matrix.majorStride))
/**
* create a tensor with a given spark Densematrix. The tensor will have the same size with
* the given spark Densematrix.
*
* @param matrix
* @return
*/
def apply(matrix: DenseMatrix): Tensor[Double] = {
val strides = if (matrix.isTransposed) {
Array(matrix.numCols, 1)
} else {
Array(1, matrix.numRows) // column major
}
apply(Storage(matrix.toArray), 1, Array(matrix.numRows, matrix.numCols), strides)
}
/**
* Create a scalar tensor of this value
* @return the created scalar tensor
*/
def scalar[T: ClassTag](value: T)(
implicit ev: TensorNumeric[T]): Tensor[T] = {
Tensor[T](Array(value), Array[Int]())
}
/**
* This is equivalent to DenseTensor.randperm[T](size)
*
* @param size
* @param ev
* @tparam T
* @return
*/
def randperm[@specialized(Float, Double) T: ClassTag](size: Int)(
implicit ev: TensorNumeric[T]): Tensor[T] = DenseTensor.randperm[T](size)
/**
* This is equivalent to tensor.expand(sizes.toArray)
*
* @param tensor
* @param sizes
* @tparam T
* @return
*/
def expand[T](tensor: Tensor[T], sizes: Int*): Tensor[T] = tensor.expand(sizes.toArray)
/**
* This is equivalent to tensor.expandAs(template)
*
* @param tensor
* @param template
* @tparam T
* @return
*/
def expandAs[T](tensor: Tensor[T], template: Tensor[T]): Tensor[T] = tensor.expandAs(template)
/**
* This is equivalent to tensor.repeatTensor(sizes.toArray)
*
* @param tensor
* @param sizes
* @tparam T
* @return
*/
def repeatTensor[T](tensor: Tensor[T], sizes: Int*): Tensor[T] =
tensor.repeatTensor(sizes.toArray)
def load[T](path : String) : Tensor[T] = {
File.load[Tensor[T]](path)
}
/**
* This is equivalent to DenseTensor.range(xmin, xmax, step)
*
* @param xmin
* @param xmax
* @param step
* @return
*/
def range[@specialized(Float, Double) T: ClassTag](xmin: Double, xmax: Double, step: Int = 1)(
implicit ev: TensorNumeric[T]): Tensor[T] = DenseTensor.range[T](xmin, xmax, step)
/**
* return a tensor of sizes filled with 1.
* @param sizes
* @return a tensor
*/
def ones[@specialized(Float, Double) T: ClassTag](sizes: Int*)(
implicit ev: TensorNumeric[T]): Tensor[T] = {
DenseTensor.ones[T](sizes.toArray)
}
/**
* Returns a 1D Gaussian kernel of size size, mean mean and standard deviation sigma.
* @param size
* @param sigma
* @param amplitude
* @param normalize
* @param mean
* @param tensor If tensor is set, will discard size, and write result to tensor.
* @return
*/
def gaussian1D[@specialized(Float, Double) T: ClassTag](
size: Int = 3,
sigma: Double = 0.25,
amplitude: Int = 1,
normalize: Boolean = false,
mean: Double = 0.5,
tensor: Tensor[T] = null)(implicit ev: TensorNumeric[T]): Tensor[T] = {
DenseTensor.gaussian1D[T](size, sigma, amplitude, normalize, mean, tensor)
}
/**
* Create a SparseTensor.
*
* @param indices dimension-D array to describe the indices of values.
* @param values non-zero values in this SparseTensor.
* @param shape shape
* @param ev
* @tparam T
* @return
*/
def sparse[T: ClassTag](
indices : Array[Array[Int]],
values : Storage[T],
shape : Array[Int])(
implicit ev: TensorNumeric[T]): Tensor[T] = {
SparseTensor(indices, values, shape, shape.length)
}
/**
* Create a SparseTensor.
*
* @param indices dimension-D array to describe the indices of values.
* @param values non-zero values in this SparseTensor.
* @param shape shape
* @param ev
* @tparam T
* @return
*/
def sparse[T: ClassTag](
indices : Array[Array[Int]],
values : Array[T],
shape : Array[Int])(
implicit ev: TensorNumeric[T]): Tensor[T] = {
sparse(indices, Storage(values), shape, shape.length)
}
/**
* Create a SparseTensor.
*
* @param indices dimension-D array to describe the indices of values.
* @param values non-zero values in this SparseTensor.
* @param shape shape
* @param dimension dimension
* @param ev
* @tparam T
* @return
*/
def sparse[T: ClassTag](
indices : Array[Array[Int]],
values : Storage[T],
shape : Array[Int],
dimension: Int)(
implicit ev: TensorNumeric[T]): Tensor[T] = {
SparseTensor(indices, values, shape, dimension)
}
/**
* Create a SparseTensor.
*
* @param indices dimension-D array to describe the indices of values.
* @param values non-zero values in this SparseTensor.
* @param shape shape
* @param dimension dimension
* @param ev
* @tparam T
* @return
*/
def sparse[T: ClassTag](
indices : Array[Array[Int]],
values : Array[T],
shape : Array[Int],
dimension: Int)(
implicit ev: TensorNumeric[T]): Tensor[T] = {
sparse(indices, Storage(values), shape, dimension)
}
/**
* Transform a DenseTensor to SparseTensor.
* @param denseTensor
* @param ev
* @tparam T
* @return
*/
def sparse[T: ClassTag](
denseTensor: Tensor[T])(implicit ev: TensorNumeric[T]): Tensor[T] = {
SparseTensor(denseTensor)
}
/**
* Create a sparse tensor with shape and number of non-zero elements.
* @param shape tensor's shape.
* @param nElement number of non-zero elements.
* @param ev
* @tparam T
* @return
*/
def sparse[T: ClassTag](
shape : Array[Int],
nElement: Int = 1)(
implicit ev: TensorNumeric[T]): Tensor[T] = {
require(nElement <= shape.product)
SparseTensor(shape, nElement)
}
/**
* Transform a sparseTensor to DenseTensor.
*
* @param sparseTensor a sparse tensor
* @param res if defined, override to res, else will generate a new tensor.
* @param ev
* @tparam T
* @return a DenseTensor.
*/
def dense[T: ClassTag](
sparseTensor: Tensor[T],
res: Tensor[T] = null)(implicit ev: TensorNumeric[T]): Tensor[T] = {
if (sparseTensor.isInstanceOf[SparseTensor[T]]) {
DenseTensor(sparseTensor.asInstanceOf[SparseTensor[T]], res)
} else if (sparseTensor.isInstanceOf[DenseTensor[T]]) {
res.copy(sparseTensor)
} else {
throw new IllegalArgumentException("Tensor.dense: Illegal tensor type.")
}
}
/**
* Concat a sequence of tensors to res tensor.
*
* @param dim concat at dim-th dimension.
* @param tensors a sequence of tensors.
* @param res result tensor.
* @param ev
* @tparam T
* @return
*/
private[bigdl] def sparseConcat[T: ClassTag](
dim: Int,
tensors: Table,
res: Tensor[T])(implicit ev: TensorNumeric[T]): Tensor[T] = {
val seqTensors = new Array[Tensor[T]](tensors.length())
var i = 0
while (i < seqTensors.length) {
seqTensors(i) = tensors[Tensor[T]](i + 1)
i += 1
}
SparseTensor.concat(dim, seqTensors, res)
}
private[bigdl] def sparseConcat[T: ClassTag](
dim: Int,
tensors: Seq[Tensor[T]],
res: Tensor[T])(implicit ev: TensorNumeric[T]): Tensor[T] = {
SparseTensor.concat(dim, tensors, res)
}
/**
* Find the distinct value and its indices in a 1D tensor.
* @param tensor a 1D tensor
* @param distinctBuffer a buffer for its distinct values.
* @param indicesBuffer a buffer for its indcies.
* @return (distinctValues, indices)
*/
def unique[T: ClassTag](
tensor: Tensor[T],
distinctBuffer: Tensor[T] = null,
indicesBuffer: Tensor[Int] = null
)(implicit ev: TensorNumeric[T]): (Tensor[T], Tensor[Int]) = {
require(tensor.isContiguous(), "unique only support contiguous tensor")
require(tensor.dim() == 1, "unique only support 1D tensor")
val array = tensor.storage().array()
val arrayOffset = tensor.storageOffset() - 1
val distinctTensor = if (null != distinctBuffer) {
distinctBuffer.resizeAs(tensor)
distinctBuffer
} else {
Tensor().resizeAs(tensor)
}
val tensorIndices = if (null != indicesBuffer) {
indicesBuffer.resizeAs(tensor)
indicesBuffer
} else {
Tensor[Int]().resizeAs(tensor)
}
val distinctValues = distinctTensor.storage().array()
val distinctValuesOffset = distinctTensor.storageOffset() - 1
val indicesArray = tensorIndices.storage().array()
val indicesOffset = tensorIndices.storageOffset() - 1
val seen = mutable.HashMap[T, Int]()
var i = 0
var nonZero = 0
while (i < tensor.nElement()) {
val x = array(i + arrayOffset)
if (!seen.contains(x)) {
distinctValues(nonZero + distinctValuesOffset) = x
seen.put(x, nonZero)
nonZero += 1
}
indicesArray(i + indicesOffset) = seen(x)
i += 1
}
// Resize distinctTensor to number of non-zero elements.
distinctTensor.resize(nonZero)
(distinctTensor, tensorIndices)
}
}
