kscience/kmath
9
0
Fork 1
Code Issues Pull Requests 6 Packages Projects Releases 16 Wiki Activity

Global refactor of tensors

Browse Source
This commit is contained in:
Alexander Nozik 2022-09-11 17:10:26 +03:00
parent b5d04ba02c
commit 20886d6f6b
No known key found for this signature in database
GPG Key ID: F7FCF2DD25C71357
8 changed files with 52 additions and 44 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
kmath-tensors/src/commonMain/kotlin/space/kscience/kmath/tensors/core/DoubleTensorAlgebra.kt
View File

@ -132,7 +132,10 @@ public open class DoubleTensorAlgebra :
val dt = asDoubleTensor()
val lastShape = shape.drop(1).toIntArray()
val newShape = if (lastShape.isNotEmpty()) lastShape else intArrayOf(1)
return DoubleTensor(newShape, dt.source.view(newShape.reduce(Int::times) * i))
return DoubleTensor(
newShape,
dt.source.view(newShape.reduce(Int::times) * i, TensorLinearStructure.linearSizeOf(newShape))
)
}
/**
@ -227,7 +230,7 @@ public open class DoubleTensorAlgebra :
override fun StructureND<Double>.minus(arg: Double): DoubleTensor = map { it - arg }
override fun StructureND<Double>.minus(arg: StructureND<Double>): DoubleTensor = zip(this, arg) { l, r -> l + r }
override fun StructureND<Double>.minus(arg: StructureND<Double>): DoubleTensor = zip(this, arg) { l, r -> l - r }
override fun Tensor<Double>.minusAssign(value: Double) {
mapInPlace { it - value }

5
kmath-tensors/src/commonMain/kotlin/space/kscience/kmath/tensors/core/IntTensorAlgebra.kt
View File

@ -221,7 +221,7 @@ public open class IntTensorAlgebra : TensorAlgebra<Int, IntRing> {
override fun StructureND<Int>.minus(arg: Int): IntTensor = map { it - arg }
override fun StructureND<Int>.minus(arg: StructureND<Int>): IntTensor = zip(this, arg) { l, r -> l + r }
override fun StructureND<Int>.minus(arg: StructureND<Int>): IntTensor = zip(this, arg) { l, r -> l - r }
override fun Tensor<Int>.minusAssign(value: Int) {
mapInPlace { it - value }
@ -283,8 +283,7 @@ public open class IntTensorAlgebra : TensorAlgebra<Int, IntRing> {
view(other.shape)
override fun StructureND<Int>.dot(other: StructureND<Int>): IntTensor {
return if (dimension in 0..2 && other.dimension in 0..2) TODO("not implemented")
else error("Only vectors and matrices are allowed in non-broadcasting dot operation")
TODO("not implemented for integers")
}
override fun diagonalEmbedding(

8
kmath-tensors/src/commonMain/kotlin/space/kscience/kmath/tensors/core/TensorLinearStructure.kt
View File

@ -15,14 +15,12 @@ import kotlin.math.max
* @param shape the shape of the tensor.
*/
public class TensorLinearStructure(override val shape: IntArray) : Strides() {
override val strides: IntArray
get() = stridesFromShape(shape)
override val strides: IntArray get() = stridesFromShape(shape)
override fun index(offset: Int): IntArray =
indexFromOffset(offset, strides, shape.size)
override val linearSize: Int
get() = shape.reduce(Int::times)
override val linearSize: Int get() = linearSizeOf(shape)
override fun equals(other: Any?): Boolean {
if (this === other) return true
@ -41,6 +39,8 @@ public class TensorLinearStructure(override val shape: IntArray) : Strides() {
public companion object {
public fun linearSizeOf(shape: IntArray): Int = shape.reduce(Int::times)
public fun stridesFromShape(shape: IntArray): IntArray {
val nDim = shape.size
val res = IntArray(nDim)

27
kmath-tensors/src/commonMain/kotlin/space/kscience/kmath/tensors/core/internal/doubleTensorHelpers.kt
View File

@ -14,11 +14,8 @@ import space.kscience.kmath.structures.DoubleBuffer
import space.kscience.kmath.structures.VirtualBuffer
import space.kscience.kmath.structures.asBuffer
import space.kscience.kmath.structures.indices
import space.kscience.kmath.tensors.core.*
import space.kscience.kmath.tensors.core.BroadcastDoubleTensorAlgebra.eye
import space.kscience.kmath.tensors.core.BufferedTensor
import space.kscience.kmath.tensors.core.DoubleTensor
import space.kscience.kmath.tensors.core.OffsetDoubleBuffer
import space.kscience.kmath.tensors.core.copyToTensor
import kotlin.math.abs
import kotlin.math.max
import kotlin.math.sqrt
@ -165,7 +162,7 @@ internal val DoubleTensor.vectors: VirtualBuffer<DoubleTensor>
return VirtualBuffer(linearSize / vectorOffset) { index ->
val offset = index * vectorOffset
DoubleTensor(vectorShape, source.view(offset))
DoubleTensor(vectorShape, source.view(offset, vectorShape.first()))
}
}
@ -174,16 +171,18 @@ internal fun DoubleTensor.vectorSequence(): Sequence<DoubleTensor> = vectors.asS
internal val DoubleTensor.matrices: VirtualBuffer<DoubleTensor>
get(){
val n = shape.size
check(n >= 2) { "Expected tensor with 2 or more dimensions, got size $n" }
val matrixOffset = shape[n - 1] * shape[n - 2]
val matrixShape = intArrayOf(shape[n - 2], shape[n - 1])
get() {
val n = shape.size
check(n >= 2) { "Expected tensor with 2 or more dimensions, got size $n" }
val matrixOffset = shape[n - 1] * shape[n - 2]
val matrixShape = intArrayOf(shape[n - 2], shape[n - 1])
return VirtualBuffer(linearSize / matrixOffset) { index ->
val offset = index * matrixOffset
DoubleTensor(matrixShape, source.view(offset))
val size = TensorLinearStructure.linearSizeOf(matrixShape)
return VirtualBuffer(linearSize / matrixOffset) { index ->
val offset = index * matrixOffset
DoubleTensor(matrixShape, source.view(offset, size))
}
}
}
internal fun DoubleTensor.matrixSequence(): Sequence<DoubleTensor> = matrices.asSequence()

2
kmath-tensors/src/commonMain/kotlin/space/kscience/kmath/tensors/core/internal/intTensorHelpers.kt
View File

@ -40,7 +40,7 @@ internal val IntTensor.vectors: VirtualBuffer<IntTensor>
return VirtualBuffer(linearSize / vectorOffset) { index ->
val offset = index * vectorOffset
IntTensor(vectorShape, source.view(offset))
IntTensor(vectorShape, source.view(offset, vectorShape.first()))
}
}

17
kmath-tensors/src/commonMain/kotlin/space/kscience/kmath/tensors/core/tensorTransform.kt
View File

@ -5,14 +5,20 @@
package space.kscience.kmath.tensors.core
import space.kscience.kmath.nd.DoubleBufferND
import space.kscience.kmath.nd.StructureND
import space.kscience.kmath.structures.DoubleBuffer
import space.kscience.kmath.structures.asBuffer
import space.kscience.kmath.tensors.api.Tensor
/**
* Create a mutable copy of given [StructureND].
*/
public fun StructureND<Double>.copyToTensor(): DoubleTensor = if (this is DoubleTensor) {
DoubleTensor(shape, source.copy())
} else if (this is DoubleBufferND && indices is TensorLinearStructure) {
DoubleTensor(shape, buffer.copy())
} else {
DoubleTensor(
shape,
@ -36,11 +42,14 @@ public fun StructureND<Int>.toDoubleTensor(): DoubleTensor {
}
/**
* Casts [Tensor] of [Double] to [DoubleTensor]
* Transforms [StructureND] of [Double] to [DoubleTensor]. Zero copy if possible, but is not guaranteed
*/
public fun StructureND<Double>.asDoubleTensor(): DoubleTensor = when (this) {
is DoubleTensor -> this
else -> copyToTensor()
public fun StructureND<Double>.asDoubleTensor(): DoubleTensor = if (this is DoubleTensor) {
this
} else if (this is DoubleBufferND && indices is TensorLinearStructure) {
DoubleTensor(shape, buffer)
} else {
copyToTensor()
}
/**

23
kmath-tensors/src/commonTest/kotlin/space/kscience/kmath/tensors/core/TestDoubleTensor.kt
View File

@ -6,10 +6,7 @@
package space.kscience.kmath.tensors.core
import space.kscience.kmath.misc.PerformancePitfall
import space.kscience.kmath.nd.DefaultStrides
import space.kscience.kmath.nd.MutableBufferND
import space.kscience.kmath.nd.as1D
import space.kscience.kmath.nd.as2D
import space.kscience.kmath.nd.*
import space.kscience.kmath.operations.invoke
import space.kscience.kmath.structures.DoubleBuffer
import space.kscience.kmath.structures.toDoubleArray
@ -66,16 +63,16 @@ internal class TestDoubleTensor {
fun testNoBufferProtocol() {
// create buffer
val doubleArray = DoubleBuffer(doubleArrayOf(1.0, 2.0, 3.0))
val doubleArray = DoubleBuffer(1.0, 2.0, 3.0)
// create ND buffers, no data is copied
val ndArray: MutableBufferND<Double> = MutableBufferND(DefaultStrides(intArrayOf(3)), doubleArray)
val ndArray: MutableBufferND<Double> = DoubleBufferND(DefaultStrides(intArrayOf(3)), doubleArray)
// map to tensors
val bufferedTensorArray = ndArray.asDoubleTensor() // strides are flipped so data copied
val tensorArray = bufferedTensorArray.asDoubleTensor() // data not contiguous so copied again
val tensorArray = ndArray.asDoubleTensor() // Data is copied because of strides change.
val tensorArrayPublic = ndArray.asDoubleTensor() // public API, data copied twice
//protective copy
val tensorArrayPublic = ndArray.copyToTensor() // public API, data copied twice
val sharedTensorArray = tensorArrayPublic.asDoubleTensor() // no data copied by matching type
assertTrue(tensorArray.source contentEquals sharedTensorArray.source)
@ -83,11 +80,11 @@ internal class TestDoubleTensor {
tensorArray[intArrayOf(0)] = 55.9
assertEquals(tensorArrayPublic[intArrayOf(0)], 1.0)
tensorArrayPublic[intArrayOf(0)] = 55.9
assertEquals(sharedTensorArray[intArrayOf(0)], 55.9)
assertEquals(bufferedTensorArray[intArrayOf(0)], 1.0)
tensorArrayPublic[intArrayOf(0)] = 57.9
assertEquals(sharedTensorArray[intArrayOf(0)], 57.9)
assertEquals(tensorArray[intArrayOf(0)], 55.9)
bufferedTensorArray[intArrayOf(0)] = 55.9
tensorArray[intArrayOf(0)] = 55.9
assertEquals(ndArray[intArrayOf(0)], 1.0)
}

7
kmath-tensors/src/commonTest/kotlin/space/kscience/kmath/tensors/core/TestDoubleTensorAlgebra.kt
View File

@ -8,6 +8,7 @@ package space.kscience.kmath.tensors.core
import space.kscience.kmath.nd.get
import space.kscience.kmath.operations.invoke
import space.kscience.kmath.testutils.assertBufferEquals
import kotlin.test.Test
import kotlin.test.assertEquals
import kotlin.test.assertFalse
@ -98,8 +99,8 @@ internal class TestDoubleTensorAlgebra {
assignResult += tensorC
assignResult += -39.4
assertTrue(expected.source contentEquals result.source)
assertTrue(expected.source contentEquals assignResult.source)
assertBufferEquals(expected.source, result.source)
assertBufferEquals(expected.source, assignResult.source)
}
@Test
@ -202,6 +203,6 @@ internal class TestDoubleTensorAlgebra {
val r = tensor.getTensor(1).map { it - 1.0 }
val res = l + r
assertTrue { intArrayOf(5, 5) contentEquals res.shape }
assertEquals(1.0, res[4, 4])
assertEquals(2.0, res[4, 4])
}
}