Семантика операций

Оптимизируйте свои подборки Сохраняйте и классифицируйте контент в соответствии со своими настройками.

Далее описывается семантика операций, определенных в интерфейсе XlaBuilder . Как правило, эти операции однозначно сопоставляются с операциями, определенными в интерфейсе RPC в xla_data.proto .

Примечание по номенклатуре: обобщенный тип данных, с которым имеет дело XLA, представляет собой N-мерный массив, содержащий элементы некоторого универсального типа (например, 32-битного числа с плавающей запятой). В документации массив используется для обозначения массива произвольной размерности. Для удобства специальные случаи имеют более конкретные и знакомые имена; например, вектор — это одномерный массив, а матрица — это двумерный массив.

После всего

См. также XlaBuilder::AfterAll .

AfterAll принимает переменное количество токенов и создает один токен. Токены — это примитивные типы, которые могут быть объединены между побочными операциями для принудительного упорядочения. AfterAll можно использовать как объединение токенов для заказа операции после набора операций.

AfterAll(operands)

Аргументы Тип Семантика
operands XlaOp переменное количество токенов

Всесобрать

См. также XlaBuilder::AllGather .

Выполняет конкатенацию между репликами.

AllGather(operand, all_gather_dim, shard_count, replica_group_ids, channel_id)

Аргументы Тип Семантика
operand XlaOp Массив для объединения реплик.
all_gather_dim int64 Размер конкатенации.
replica_groups вектор векторов int64 Группы, между которыми выполняется конкатенация.
channel_id необязательный int64 Необязательный идентификатор канала для связи между модулями.
  • replica_groups — это список групп реплик, между которыми выполняется конкатенация (идентификатор реплики для текущей реплики можно получить с помощью ReplicaId ). Порядок реплик в каждой группе определяет порядок расположения их входов в результате. replica_groups должен быть либо пустым (в этом случае все реплики принадлежат одной группе, упорядоченной от 0 до N - 1 ), либо содержать такое же количество элементов, как и количество реплик. Например, replica_groups = {0, 2}, {1, 3} выполняет конкатенацию между репликами 0 и 2 , а также 1 и 3 .
  • shard_count — это размер каждой группы реплик. Это нужно в тех случаях, когда replica_groups пусты.
  • channel_id используется для связи между модулями: только операции all-gather с одним и тем же channel_id могут взаимодействовать друг с другом.

Выходная форма — это входная форма, в all_gather_dim увеличено в shard_count . Например, если есть две реплики и операнд имеет значение [1.0, 2.5] и [3.0, 5.25] соответственно на двух репликах, то выходное значение этой операции, где all_gather_dim равно 0 , будет [1.0, 2.5, 3.0, 5.25] на обеих репликах.

ВсеСвернуть

См. также XlaBuilder::AllReduce .

Выполняет настраиваемые вычисления между репликами.

AllReduce(operand, computation, replica_group_ids, channel_id)

Аргументы Тип Семантика
operand XlaOp Массив или непустой кортеж массивов для сокращения реплик.
computation XlaComputation Вычисление сокращения
replica_groups вектор векторов int64 Группы, между которыми выполняются сокращения
channel_id необязательный int64 Дополнительный идентификатор канала для межмодульной связи
  • Когда operand представляет собой кортеж массивов, все сведение выполняется к каждому элементу кортежа.
  • replica_groups — это список групп реплик, между которыми выполняется сокращение (идентификатор реплики для текущей реплики можно получить с помощью ReplicaId ). replica_groups должен быть либо пустым (в этом случае все реплики принадлежат одной группе), либо содержать такое же количество элементов, как и количество реплик. Например, replica_groups = {0, 2}, {1, 3} выполняет сокращение между репликами 0 и 2 , а также 1 и 3 .
  • channel_id используется для связи между модулями: только операции all-reduce с одним и тем же channel_id могут взаимодействовать друг с другом.

Выходная форма такая же, как и входная форма. Например, если есть две реплики и операнд имеет значение [1.0, 2.5] и [3.0, 5.25] соответственно на двух репликах, то выходное значение этой операции и вычисления суммирования будет [4.0, 7.75] на обеих репликах. реплики. Если ввод представляет собой кортеж, вывод также является кортежем.

Вычисление результата AllReduce требует наличия одного входа от каждой реплики, поэтому, если одна реплика выполняет узел AllReduce больше раз, чем другая, первая реплика будет ждать вечно. Поскольку на всех репликах работает одна и та же программа, вариантов для этого не так много, но это возможно, когда условие цикла while зависит от данных из потока, а данные, которые загружаются, заставляют цикл while повторяться больше раз. на одной реплике, чем на другой.

ВсеВсем

См. также XlaBuilder::AllToAll .

AllToAll — это коллективная операция, которая отправляет данные со всех ядер на все ядра. Он имеет две фазы:

  1. Фаза рассеяния. На каждом ядре операнд разбивается на количество блоков split_count по split_dimensions , и блоки распределяются по всем ядрам, например, i-й блок отправляется i-му ядру.
  2. Фаза сбора. Каждое ядро ​​объединяет полученные блоки по параметру concat_dimension .

Участвующие ядра можно настроить с помощью:

  • replica_groups : каждая ReplicaGroup содержит список идентификаторов реплик, участвующих в вычислении (идентификатор реплики для текущей реплики можно получить с помощью ReplicaId ). AllToAll будет применяться внутри подгрупп в указанном порядке. Например, replica_groups = { {1,2,3}, {4,5,0} } означает, что AllToAll будет применяться в репликах {1, 2, 3} и на этапе сбора, а полученные блоки будут быть объединены в том же порядке 1, 2, 3. Затем к репликам 4, 5, 0 будет применен другой AllToAll, и порядок объединения также будет 4, 5, 0. Если replica_groups пуст, все реплики принадлежат одному группы в порядке их появления.

Предпосылки:

  • Размер размерности операнда в split_dimension делится на split_count .
  • Форма операнда не кортеж.

AllToAll(operand, split_dimension, concat_dimension, split_count, replica_groups)

Аргументы Тип Семантика
operand XlaOp n-мерный входной массив
split_dimension int64 Значение в интервале [0, n) , обозначающее измерение, по которому разбивается операнд.
concat_dimension int64 значение в интервале [0, n) , обозначающее измерение, по которому объединяются разделяемые блоки.
split_count int64 количество ядер, участвующих в этой операции. Если replica_groups пуст, это должно быть количество реплик; в противном случае это должно быть равно количеству реплик в каждой группе.
replica_groups ReplicaGroup каждая группа содержит список идентификаторов реплик.

Ниже показан пример Alltoall.

XlaBuilder b("alltoall");
auto x = Parameter(&b, 0, ShapeUtil::MakeShape(F32, {4, 16}), "x");
AllToAll(x, /*split_dimension=*/1, /*concat_dimension=*/0, /*split_count=*/4);

В этом примере в Alltoall участвуют 4 ядра. На каждом ядре операнд разбивается на 4 части по размерности 0, поэтому каждая часть имеет форму f32[4,4]. 4 части разбросаны по всем ядрам. Затем каждое ядро ​​объединяет полученные части по измерению 1 в порядке ядра 0-4. Таким образом, выход каждого ядра имеет вид f32[16,4].

ПакетНормаГрад

Подробное описание алгоритма см. также XlaBuilder::BatchNormGrad и исходном документе по нормализации пакетов.

Вычисляет градиенты нормы партии.

BatchNormGrad(operand, scale, mean, variance, grad_output, epsilon, feature_index)

Аргументы Тип Семантика
operand XlaOp n-мерный массив для нормализации (x)
scale XlaOp 1-мерный массив (\(\gamma\))
mean XlaOp 1-мерный массив (\(\mu\))
variance XlaOp 1-мерный массив (\(\sigma^2\))
grad_output XlaOp Градиенты, переданные в BatchNormTraining (\( \nabla y\))
epsilon float Значение эпсилон (\(\epsilon\))
feature_index int64 Индекс для измерения функции в operand

Для каждого объекта в измерении объекта ( feature_index — это индекс измерения объекта в operand ), операция вычисляет градиенты относительно operand , offset и scale по всем другим измерениям. feature_index должен быть допустимым индексом для измерения функции в operand .

Три градиента определяются следующими формулами (при условии, что 4-мерный массив является operand и имеет индекс размерности объекта l , размер пакета m и пространственные размеры w и h ):

\[ \begin{split} c_l&= \frac{1}{mwh}\sum_{i=1}^m\sum_{j=1}^w\sum_{k=1}^h \left( \nabla y_{ijkl} \frac{x_{ijkl} - \mu_l}{\sigma^2_l+\epsilon} \right) \\\\ d_l&= \frac{1}{mwh}\sum_{i=1}^m\sum_{j=1}^w\sum_{k=1}^h \nabla y_{ijkl} \\\\ \nabla x_{ijkl} &= \frac{\gamma_{l} }{\sqrt{\sigma^2_{l}+\epsilon} } \left( \nabla y_{ijkl} - d_l - c_l (x_{ijkl} - \mu_{l}) \right) \\\\ \nabla \gamma_l &= \sum_{i=1}^m\sum_{j=1}^w\sum_{k=1}^h \left( \nabla y_{ijkl} \frac{x_{ijkl} - \mu_l}{\sqrt{\sigma^2_{l}+\epsilon} } \right) \\\\\ \nabla \beta_l &= \sum_{i=1}^m\sum_{j=1}^w\sum_{k=1}^h \nabla y_{ijkl} \end{split} \]

mean входных данных и variance представляют собой значения моментов в пакетных и пространственных измерениях.

Выходной тип представляет собой кортеж из трех дескрипторов:

Выходы Тип Семантика
grad_operand XlaOp градиент относительно входного operand (\( \nabla x\))
grad_scale XlaOp градиент по отношению к scale ввода (\( \nabla \gamma\))
grad_offset XlaOp градиент относительно входного offset (\( \nabla \beta\))

BatchNormInference

Подробное описание алгоритма см. также XlaBuilder::BatchNormInference и исходном документе по нормализации пакетов.

Нормализует массив по пакетным и пространственным измерениям.

BatchNormInference(operand, scale, offset, mean, variance, epsilon, feature_index)

Аргументы Тип Семантика
operand XlaOp n-мерный массив для нормализации
scale XlaOp 1-мерный массив
offset XlaOp 1-мерный массив
mean XlaOp 1-мерный массив
variance XlaOp 1-мерный массив
epsilon float Значение эпсилон
feature_index int64 Индекс для измерения функции в operand

Для каждого признака в измерении признаков ( feature_index — это индекс измерения признаков в operand ) операция вычисляет среднее значение и дисперсию по всем другим измерениям и использует среднее значение и дисперсию для нормализации каждого элемента в operand . feature_index должен быть допустимым индексом для измерения функции в operand .

BatchNormInference эквивалентен вызову BatchNormTraining без вычисления mean и variance для каждой партии. Вместо этого он использует входное mean и variance в качестве оценочных значений. Цель этой операции — уменьшить задержку при выводе, отсюда и название BatchNormInference .

Выход представляет собой n-мерный нормализованный массив той же формы, что и входной operand .

ПакетНормаОбучение

Подробное описание алгоритма см. также XlaBuilder::BatchNormTraining и the original batch normalization paper пакетов.

Нормализует массив по пакетным и пространственным измерениям.

BatchNormTraining(operand, scale, offset, epsilon, feature_index)

Аргументы Тип Семантика
operand XlaOp n-мерный массив для нормализации (x)
scale XlaOp 1-мерный массив (\(\gamma\))
offset XlaOp 1-мерный массив (\(\beta\))
epsilon float Значение эпсилон (\(\epsilon\))
feature_index int64 Индекс для измерения функции в operand

Для каждого признака в измерении признаков ( feature_index — это индекс измерения признаков в operand ) операция вычисляет среднее значение и дисперсию по всем другим измерениям и использует среднее значение и дисперсию для нормализации каждого элемента в operand . feature_index должен быть допустимым индексом для измерения функции в operand .

Алгоритм работает следующим образом для каждого пакета в operand \(x\) , который содержит m элементов с w и h в качестве размера пространственных измерений (при условии, что operand представляет собой четырехмерный массив):

  • Вычисляет пакетное среднее \(\mu_l\) для каждой функции l в измерении функции:\(\mu_l=\frac{1}{mwh}\sum_{i=1}^m\sum_{j=1}^w\sum_{k=1}^h x_{ijkl}\)

  • Вычисляет отклонение партии \(\sigma^2_l\):\(\sigma^2_l=\frac{1}{mwh}\sum_{i=1}^m\sum_{j=1}^w\sum_{k=1}^h (x_{ijkl} - \mu_l)^2\)

  • Нормализует, масштабирует и сдвигает:\(y_{ijkl}=\frac{\gamma_l(x_{ijkl}-\mu_l)}{\sqrt[2]{\sigma^2_l+\epsilon} }+\beta_l\)

Значение эпсилон, обычно небольшое число, добавляется, чтобы избежать ошибок деления на ноль.

Выходной тип представляет собой кортеж из трех XlaOp s:

Выходы Тип Семантика
output XlaOp n-мерный массив той же формы, что и входной operand (y)
batch_mean XlaOp 1-мерный массив (\(\mu\))
batch_var XlaOp 1-мерный массив (\(\sigma^2\))

batch_mean и batch_var — это моменты, рассчитанные по пакетному и пространственному измерениям с использованием приведенных выше формул.

BitcastConvertType

См. также XlaBuilder::BitcastConvertType .

Подобно tf.bitcast в TensorFlow, выполняет поэлементную операцию побитовой передачи из формы данных в целевую форму. Размер входных и выходных данных должен совпадать: например, элементы s32 становятся элементами f32 посредством процедуры побитовой передачи, а один элемент s32 становится четырьмя элементами s8 . Bitcast реализован как низкоуровневое приведение, поэтому машины с разными представлениями с плавающей запятой будут давать разные результаты.

BitcastConvertType(operand, new_element_type)

Аргументы Тип Семантика
operand XlaOp массив типа T с размерами D
new_element_type PrimitiveType тип U

Размеры операнда и целевой формы должны совпадать, за исключением последнего измерения, которое изменится в соотношении размера примитива до и после преобразования.

Типы элементов источника и назначения не должны быть кортежами.

Bitcast-преобразование в примитивный тип другой ширины

BitcastConvert поддерживает случай, когда размер выходного элемента типа T' не равен размеру входного элемента T . Поскольку вся операция концептуально представляет собой побитовое преобразование и не изменяет базовые байты, форма выходного элемента должна измениться. Для B = sizeof(T), B' = sizeof(T') возможны два случая.

Во-первых, когда B > B' , выходная форма получает новое наименьшее измерение размера B/B' . Например:

  f16[10,2]{1,0} %output = f16[10,2]{1,0} bitcast-convert(f32[10]{0} %input)

Правило остается тем же для эффективных скаляров:

  f16[2]{0} %output = f16[2]{0} bitcast-convert(f32[] %input)

В качестве альтернативы, для B' > B инструкция требует, чтобы последнее логическое измерение входной формы было равно B'/B , и это измерение отбрасывается во время преобразования:

  f32[10]{0} %output = f32[10]{0} bitcast-convert(f16[10,2]{1,0} %input)

Обратите внимание, что преобразования между разными разрядностями не являются поэлементными.

Транслировать

См. также XlaBuilder::Broadcast .

Добавляет измерения в массив, дублируя данные в массиве.

Broadcast(operand, broadcast_sizes)

Аргументы Тип Семантика
operand XlaOp Массив для дублирования
broadcast_sizes ArraySlice<int64> Размеры новых габаритов

Новые размеры вставляются слева, т. е. если broadcast_sizes имеют значения {a0, ..., aN} а форма операнда имеет размеры {b0, ..., bM} то форма вывода имеет размеры {a0, ..., aN, b0, ..., bM} .

Новые размерности индексируются копиями операнда, т.е.

output[i0, ..., iN, j0, ..., jM] = operand[j0, ..., jM]

Например, если operand является скаляром f32 со значением 2.0f , а broadcast_sizes равны {2, 3} , тогда результатом будет массив формы f32[2, 3] и все значения в результате будут 2.0f .

BroadcastInDim

См. также XlaBuilder::BroadcastInDim .

Увеличивает размер и ранг массива за счет дублирования данных в массиве.

BroadcastInDim(operand, out_dim_size, broadcast_dimensions)

Аргументы Тип Семантика
operand XlaOp Массив для дублирования
out_dim_size ArraySlice<int64> Размеры габаритов целевой формы
broadcast_dimensions ArraySlice<int64> Какому измерению в целевой форме соответствует каждое измерение формы операнда

Аналогичен Broadcast, но позволяет добавлять размеры в любом месте и расширять существующие размеры с размером 1.

operand транслируется в форму, описанную out_dim_size . broadcast_dimensions сопоставляет размерность operand с размерностью целевой формы, т. е. i-е измерение операнда сопоставляется с Broad_dimension[i]-м измерением выходной фигуры. Размеры operand должны иметь размер 1 или быть того же размера, что и размер выходной формы, с которой они сопоставлены. Остальные измерения заполняются измерениями размера 1. Затем вещание с вырожденным измерением распространяется по этим вырожденным измерениям, чтобы достичь выходной формы. Семантика подробно описана на странице трансляции .

Вызов

См. также XlaBuilder::Call .

Вызывает вычисление с заданными аргументами.

Call(computation, args...)

Аргументы Тип Семантика
computation XlaComputation вычисление типа T_0, T_1, ..., T_{N-1} -> S с N параметрами произвольного типа
args последовательность N XlaOp s N аргументов произвольного типа

Арность и типы args должны соответствовать параметрам computation . Допускается отсутствие args .

Холецкий

См. также XlaBuilder::Cholesky .

Вычисляет разложение Холецкого пакета симметричных (эрмитовых) положительно определенных матриц.

Cholesky(a, lower)

Аргументы Тип Семантика
a XlaOp массив ранга > 2 сложного типа или типа с плавающей запятой.
lower bool следует ли использовать верхний или нижний треугольник a .

Если lower равно true , вычисляет нижние треугольные матрицы l такие, что \( a = l . l^T \). Если lower равен false , вычисляет верхнетреугольные матрицы u такие, что \( a = u^T . u \).

Входные данные считываются только из нижнего/верхнего треугольника a , в зависимости от значения lower . Значения из другого треугольника игнорируются. Выходные данные возвращаются в том же треугольнике; значения в другом треугольнике определяются реализацией и могут быть любыми.

Если ранг a больше 2, a рассматривается как пакет матриц, где все измерения, кроме младших 2, являются размерами пакета.

Если a не является симметричным (эрмитовым) положительно определенным, результат определяется реализацией.

Зажим

См. также XlaBuilder::Clamp .

Фиксирует операнд в пределах диапазона между минимальным и максимальным значением.

Clamp(min, operand, max)

Аргументы Тип Семантика
min XlaOp массив типа Т
operand XlaOp массив типа Т
max XlaOp массив типа Т

Учитывая операнд и минимальное и максимальное значения, возвращает операнд, если он находится в диапазоне между минимальным и максимальным, иначе возвращает минимальное значение, если операнд ниже этого диапазона, или максимальное значение, если операнд выше этого диапазона. То есть clamp(a, x, b) = min(max(a, x), b) .

Все три массива должны быть одинаковой формы. В качестве альтернативы, в качестве ограниченной формы широковещательной передачи min и/или max могут быть скаляром типа T .

Пример со скалярным min и max :

let operand: s32[3] = {-1, 5, 9};
let min: s32 = 0;
let max: s32 = 6;
==>
Clamp(min, operand, max) = s32[3]{0, 5, 6};

Крах

См. также XlaBuilder::Collapse и операцию tf.reshape .

Сворачивает измерения массива в одно измерение.

Collapse(operand, dimensions)

Аргументы Тип Семантика
operand XlaOp массив типа Т
dimensions int64 вектор упорядоченное, последовательное подмножество измерений T.

Collapse заменяет данное подмножество измерений операнда одним измерением. Входные аргументы представляют собой произвольный массив типа T и постоянный во время компиляции вектор индексов размерности. Индексы размерности должны быть упорядоченными (числа размерности от младших к старшим), последовательным подмножеством измерений T. Таким образом, {0, 1, 2}, {0, 1} или {1, 2} являются допустимыми наборами измерений, а {1, 0} или {0, 2} — нет. Они заменяются одним новым размером, находящимся в той же позиции в последовательности размеров, что и те, которые они заменяют, с новым размером размера, равным произведению исходных размеров размеров. Наименьшее число измерений в dimensions является самым медленным изменяющимся измерением (наиболее важным) в гнезде цикла, которое сворачивает эти измерения, а наибольшее число измерений изменяется быстрее всего (наиболее второстепенное). См. оператор tf.reshape , если требуется более общий порядок свертывания.

Например, пусть v будет массивом из 24 элементов:

let v = f32[4x2x3] { { {10, 11, 12},  {15, 16, 17} },
{ {20, 21, 22},  {25, 26, 27} },
{ {30, 31, 32},  {35, 36, 37} },
{ {40, 41, 42},  {45, 46, 47} } };

// Collapse to a single dimension, leaving one dimension.
let v012 = Collapse(v, {0,1,2});
then v012 == f32[24] {10, 11, 12, 15, 16, 17,
20, 21, 22, 25, 26, 27,
30, 31, 32, 35, 36, 37,
40, 41, 42, 45, 46, 47};

// Collapse the two lower dimensions, leaving two dimensions.
let v01 = Collapse(v, {0,1});
then v01 == f32[4x6] { {10, 11, 12, 15, 16, 17},
{20, 21, 22, 25, 26, 27},
{30, 31, 32, 35, 36, 37},
{40, 41, 42, 45, 46, 47} };

// Collapse the two higher dimensions, leaving two dimensions.
let v12 = Collapse(v, {1,2});
then v12 == f32[8x3] { {10, 11, 12},
{15, 16, 17},
{20, 21, 22},
{25, 26, 27},
{30, 31, 32},
{35, 36, 37},
{40, 41, 42},
{45, 46, 47} };

КоллективPermute

См. также XlaBuilder::CollectivePermute .

CollectivePermute — это коллективная операция, которая отправляет и получает перекрестные реплики данных.

CollectivePermute(operand, source_target_pairs)

Аргументы Тип Семантика
operand XlaOp n-мерный входной массив
source_target_pairs <int64, int64> вектор Список пар (source_replica_id, target_replica_id). Для каждой пары операнд отправляется от исходной реплики к целевой реплике.

Обратите внимание, что существуют следующие ограничения на source_target_pair :

  • Любые две пары не должны иметь одинаковый идентификатор целевой реплики, и они не должны иметь одинаковый идентификатор исходной реплики.
  • Если идентификатор реплики не является целью ни в одной паре, то выход этой реплики представляет собой тензор, состоящий из 0 (ов) той же формы, что и вход.

Объединить

См. также XlaBuilder::ConcatInDim .

Concatenate составляет массив из нескольких операндов массива. Массив имеет тот же ранг, что и каждый из входных операндов массива (которые должны иметь тот же ранг, что и каждый другой), и содержит аргументы в том порядке, в котором они были указаны.

Concatenate(operands..., dimension)

Аргументы Тип Семантика
operands последовательность N XlaOp N массивов типа T размерностью [L0, L1, ...]. Требуется N >= 1.
dimension int64 Значение в интервале [0, N) , обозначающее измерение, которое будет конкатенировано между operands .

За исключением dimension , все размеры должны быть одинаковыми. Это связано с тем, что XLA не поддерживает "рваные" массивы. Также обратите внимание, что значения ранга-0 не могут быть объединены (поскольку невозможно назвать измерение, по которому происходит объединение).

1-мерный пример:

Concat({ {2, 3}, {4, 5}, {6, 7} }, 0)
>>> {2, 3, 4, 5, 6, 7}

2-мерный пример:

let a = {
{1, 2},
{3, 4},
{5, 6},
};
let b = {
{7, 8},
};
Concat({a, b}, 0)
>>> {
{1, 2},
{3, 4},
{5, 6},
{7, 8},
}

Диаграмма:

Условный

См. также XlaBuilder::Conditional .

Conditional(pred, true_operand, true_computation, false_operand, false_computation)

Аргументы Тип Семантика
pred XlaOp Скаляр типа PRED
true_operand XlaOp Аргумент типа \(T_0\)
true_computation XlaComputation XlaComputation типа \(T_0 \to S\)
false_operand XlaOp Аргумент типа \(T_1\)
false_computation XlaComputation XlaComputation типа \(T_1 \to S\)

Выполняет true_computation , если pred равно true , false_computation , если pred равно false , и возвращает результат.

true_computation должен принимать один аргумент типа \(T_0\) и будет вызываться с true_operand , который должен быть того же типа. false_computation должно принимать один аргумент типа \(T_1\) и будет вызываться с false_operand , который должен быть того же типа. Тип возвращаемого значения true_computation и false_computation должен быть одинаковым.

Обратите внимание, что только одно из true_computation и false_computation будет выполнено в зависимости от значения pred .

Conditional(branch_index, branch_computations, branch_operands)

Аргументы Тип Семантика
branch_index XlaOp Скаляр типа S32
branch_computations последовательность N XlaComputation XlaComputations типа \( T_0 \to S , T_1 \to S , ..., T_{N-1} \to S \)
branch_operands последовательность N XlaOp Аргументы типа \( T_0 , T_1 , ..., T_{N-1} \)

Выполняет branch_computations[branch_index] и возвращает результат. Если branch_index — это S32 , который меньше 0 или >= N, то branch_computations[N-1] выполняется как ветвь по умолчанию.

Каждый branch_computations[b] должен принимать один аргумент типа T_b и будет вызываться с branch_operands[b] , которые должны быть того же типа. Тип возвращаемого значения каждой branch_computations[b] должен быть одинаковым.

Обратите внимание, что только один из branch_computations будет выполнен в зависимости от значения branch_index .

Конв (свертка)

См. также XlaBuilder::Conv .

То же, что и ConvWithGeneralPadding, но заполнение указывается сокращенно как SAME или VALID. SAME padding дополняет ввод ( lhs ) нулями, чтобы вывод имел ту же форму, что и ввод, если не принимать во внимание шаг. ДЕЙСТВИТЕЛЬНОЕ заполнение просто означает отсутствие заполнения.

ConvWithGeneralPadding (свертка)

См. также XlaBuilder::ConvWithGeneralPadding .

Вычисляет свертки типа используемых в нейронных сетях. Здесь свертка может рассматриваться как n-мерное окно, перемещающееся по n-мерной базовой области, и вычисление выполняется для каждой возможной позиции окна.

Аргументы Тип Семантика
lhs XlaOp ранг n+2 массив входов
rhs XlaOp rank n+2 массив весов ядра
window_strides ArraySlice<int64> й массив шагов ядра
padding ArraySlice< pair<int64, int64>> й массив (низкий, высокий) заполнения
lhs_dilation ArraySlice<int64> массив коэффициентов дилатации nd lhs
rhs_dilation ArraySlice<int64> массив коэффициентов дилатации nd rhs
feature_group_count int64 количество групп функций
batch_group_count int64 количество пакетных групп

Пусть n будет количеством пространственных измерений. Аргумент lhs представляет собой массив рангов n+2, описывающий базовую область. Это называется входом, хотя, конечно, правая сторона тоже является входом. В нейронной сети это входные активации. n + 2 измерения в следующем порядке:

  • batch : Каждая координата в этом измерении представляет собой независимый ввод, для которого выполняется свертка.
  • z/depth/features : каждая (y,x) позиция в базовой области имеет связанный с ней вектор, который входит в это измерение.
  • spatial_dims : описывает n пространственных измерений, определяющих базовую область, по которой перемещается окно.

Аргумент rhs представляет собой массив рангов n+2, описывающий сверточный фильтр/ядро/окно. Размеры в таком порядке:

  • output-z : z -размер вывода.
  • input-z : размер этого измерения, умноженный на feature_group_count , должен равняться размеру измерения z в левой части.
  • spatial_dims : описывает n пространственных измерений, которые определяют nd окно, которое перемещается по базовой области.

Аргумент window_strides задает шаг сверточного окна в пространственных измерениях. Например, если шаг в первом пространственном измерении равен 3, то окно можно разместить только в координатах, где первый пространственный индекс делится на 3.

Аргумент padding указывает количество нулевого заполнения, которое должно быть применено к базовой области. Величина заполнения может быть отрицательной — абсолютное значение отрицательного заполнения указывает количество элементов, которые необходимо удалить из указанного измерения перед выполнением свертки. padding[0] задает заполнение для измерения y , а padding[1] задает заполнение для измерения x . Каждая пара имеет нижнее заполнение в качестве первого элемента и высокое заполнение в качестве второго элемента. Низкое заполнение применяется в направлении более низких индексов, а высокое заполнение применяется в направлении более высоких индексов. Например, если padding[1] равен (2,3) , то во втором пространственном измерении будет добавлено 2 нуля слева и 3 нуля справа. Использование заполнения эквивалентно вставке тех же нулевых значений во входные данные ( lhs ) перед выполнением свертки.

lhs_dilation и rhs_dilation задают коэффициент расширения, который будет применяться к левой и правой сторонам соответственно в каждом пространственном измерении. Если коэффициент расширения в пространственном измерении равен d, то между каждым из элементов в этом измерении неявно размещается d-1 дыр, увеличивая размер массива. Отверстия заполнены неактивным значением, которое для свертки означает нули.

Расширение правой извилины также называют атрофической извилиной. Для получения дополнительной информации см. tf.nn.atrous_conv2d . Расширение левой стороны также называют транспонированной сверток. Для получения дополнительной информации см. tf.nn.conv2d_transpose .

Аргумент feature_group_count (значение по умолчанию 1) можно использовать для сгруппированных сверток. feature_group_count должен быть делителем как входного, так и выходного измерения объекта. Если feature_group_count больше 1, это означает, что концептуально размерность входных и выходных объектов, а также rhs выходная размерность объектов равномерно разделены на множество групп feature_group_count , каждая из которых состоит из последовательной подпоследовательности объектов. Размерность входного объекта rhs поля должна быть равна lhs входного объекта левого края, деленному на feature_group_count (поэтому он уже имеет размер группы входных объектов). i-е группы используются вместе для вычисления feature_group_count многих отдельных сверток. Результаты этих сверток объединяются вместе в измерении выходных объектов.

Для глубинной свертки аргумент feature_group_count будет установлен на входное измерение объекта, а форма фильтра будет изменена с [filter_height, filter_width, in_channels, channel_multiplier] на [filter_height, filter_width, 1, in_channels * channel_multiplier] . Для получения дополнительной информации см. tf.nn.depthwise_conv2d .

batch_group_count (значение по умолчанию 1) можно использовать для сгруппированных фильтров во время обратного распространения. batch_group_count должен быть делителем размера lhs (входного) пакетного измерения. Если batch_group_count больше 1, это означает, что размер выходного пакета должен быть равен размеру input batch / batch_group_count . batch_group_count должен быть делителем размера выходного объекта.

Выходная форма имеет следующие размеры в следующем порядке:

  • batch : размер этого измерения, умноженный batch_group_count , должен равняться размеру batch измерения в левых столбцах.
  • z : Тот же размер, что и output-z в ядре ( rhs ).
  • spatial_dims : одно значение для каждого допустимого размещения сверточного окна.

На рисунке выше показано, как работает поле batch_group_count . По сути, мы разбиваем каждую партию batch_group_count на группы batch_group_count и делаем то же самое для выходных объектов. Затем для каждой из этих групп мы делаем попарные свертки и объединяем выходные данные по измерению выходных признаков. Операционная семантика всех остальных измерений (характерных и пространственных) остается прежней.

Допустимые размещения сверточных окон определяются шагами и размером базовой области после заполнения.

Чтобы описать, что делает свертка, рассмотрим двумерную свертка и выберите в выходных данных некоторую фиксированную batch , координаты z , y , x . Тогда (y,x) — это положение угла окна в базовой области (например, верхний левый угол, в зависимости от того, как вы интерпретируете пространственные измерения). Теперь у нас есть 2D-окно, взятое из базовой области, где каждая 2D-точка связана с 1D-вектором, поэтому мы получаем 3D-блок. Из сверточного ядра, поскольку мы зафиксировали выходную координату z , у нас также есть 3d-блок. Два прямоугольника имеют одинаковые размеры, поэтому мы можем взять сумму поэлементных произведений между двумя прямоугольниками (аналогично скалярному произведению). Это выходное значение.

Обратите внимание, что если output-z равен, например, 5, то каждая позиция окна создает 5 значений на выходе в z -измерении вывода. Эти значения различаются тем, какая часть ядра свертки используется — для каждой output-z используется отдельный трехмерный блок значений. Таким образом, вы можете представить это как 5 отдельных сверток с разными фильтрами для каждой из них.

Вот псевдокод для двумерной свертки с отступами и шагами:

for (b, oz, oy, ox) {  // output coordinates
  value = 0;
  for (iz, ky, kx) {  // kernel coordinates and input z
    iy = oy*stride_y + ky - pad_low_y;
    ix = ox*stride_x + kx - pad_low_x;
    if ((iy, ix) inside the base area considered without padding) {
      value += input(b, iz, iy, ix) * kernel(oz, iz, ky, kx);
    }
  }
  output(b, oz, oy, ox) = value;
}

КонвертироватьЭлементТип

См. также XlaBuilder::ConvertElementType .

Подобно поэлементному static_cast в C++, выполняет поэлементную операцию преобразования формы данных в целевую форму. Размеры должны совпадать, а преобразование — поэлементное; например, элементы s32 становятся элементами f32 посредством процедуры преобразования s32 в f32 .

ConvertElementType(operand, new_element_type)

Аргументы Тип Семантика
operand XlaOp массив типа T с размерами D
new_element_type PrimitiveType тип U

Размеры операнда и целевой формы должны совпадать. Типы элементов источника и назначения не должны быть кортежами.

Преобразование, такое как T=s32 в U=f32 , будет выполнять нормализующую процедуру преобразования int-to-float, такую ​​как округление до ближайшего четного.

let a: s32[3] = {0, 1, 2};
let b: f32[3] = convert(a, f32);
then b == f32[3]{0.0, 1.0, 2.0}

CrossReplicaSum

Выполняет AllReduce с вычислением суммирования.

Пользовательский вызов

См. также XlaBuilder::CustomCall .

Вызов предоставленной пользователем функции в рамках вычисления.

CustomCall(target_name, args..., shape)

Аргументы Тип Семантика
target_name string Имя функции. Будет выдана команда вызова, нацеленная на это имя символа.
args последовательность N XlaOp s N аргументов произвольного типа, которые будут переданы в функцию.
shape Shape Выходная форма функции

Сигнатура функции одинакова, независимо от арности или типа аргументов:

extern "C" void target_name(void* out, void** in);

Например, если CustomCall используется следующим образом:

let x = f32[2] {1,2};
let y = f32[2x3] { {10, 20, 30}, {40, 50, 60} };

CustomCall("myfunc", {x, y}, f32[3x3])

Вот пример реализации myfunc :

extern "C" void myfunc(void* out, void** in) {
  float (&x)[2] = *static_cast<float(*)[2]>(in[0]);
  float (&y)[2][3] = *static_cast<float(*)[2][3]>(in[1]);
  EXPECT_EQ(1, x[0]);
  EXPECT_EQ(2, x[1]);
  EXPECT_EQ(10, y[0][0]);
  EXPECT_EQ(20, y[0][1]);
  EXPECT_EQ(30, y[0][2]);
  EXPECT_EQ(40, y[1][0]);
  EXPECT_EQ(50, y[1][1]);
  EXPECT_EQ(60, y[1][2]);
  float (&z)[3][3] = *static_cast<float(*)[3][3]>(out);
  z[0][0] = x[1] + y[1][0];
  // ...
}

Пользовательская функция не должна иметь побочных эффектов, а ее выполнение должно быть идемпотентным.

точка

См. также XlaBuilder::Dot .

Dot(lhs, rhs)

Аргументы Тип Семантика
lhs XlaOp массив типа Т
rhs XlaOp массив типа Т

Точная семантика этой операции зависит от рангов операндов:

Вход Вывод Семантика
вектор [n] dot вектор [n] скаляр вектор скалярное произведение
матрица [mxk] dot вектор [k] вектор [м] умножение матрицы на вектор
матрица [mxk] dot матрица [kxn] матрица [mxn] умножение матрицы на матрицу

Операция выполняет сумму произведений второго измерения lhs (или первого, если оно имеет ранг 1) и первого измерения rhs . Это "сокращенные" размеры. Сокращенные размеры lhs и rhs стороны должны быть одного размера. На практике его можно использовать для выполнения скалярных произведений между векторами, векторных/матричных умножений или матричных/матричных умножений.

ДотГенерал

См. также XlaBuilder::DotGeneral .

DotGeneral(lhs, rhs, dimension_numbers)

Аргументы Тип Семантика
lhs XlaOp массив типа Т
rhs XlaOp массив типа Т
dimension_numbers DotDimensionNumbers контрактные и размерные номера партии

То же, что и точка, но позволяет указывать номера размеров контракта и партии как для «левого», так и для «правого» столбца.

Поля DotDimensionNumbers Тип Семантика
'lhs_contracting_dimensions' повтор int64 'lhs' сокращающиеся размерные числа
'rhs_contracting_dimensions' повтор int64 'RHS' цифры размера сужения
'lhs_batch_dimensions' повтор int64 Номера размеров партии 'lhs'
'rhs_batch_dimensions' повтор int64 номера размеров партии 'правая сторона'

DotGeneral выполняет суммирование продуктов по сокращающимся размерам, указанным в 'dimension_numbers'.

Связанные сокращающиеся размерные числа из «левого» и «правого» не обязательно должны быть одинаковыми, но должны иметь одинаковые размерные размеры.

Пример с сокращающимися размерными числами:

lhs = { {1.0, 2.0, 3.0},
{4.0, 5.0, 6.0} }

rhs = { {1.0, 1.0, 1.0},
{2.0, 2.0, 2.0} }

DotDimensionNumbers dnums;
dnums.add_lhs_contracting_dimensions(1);
dnums.add_rhs_contracting_dimensions(1);

DotGeneral(lhs, rhs, dnums) -> { {6.0, 12.0},
{15.0, 30.0} }

Связанные номера размеров партии из 'lhs' и 'rhs' должны иметь одинаковые размеры размеров.

Пример с номерами размеров партии (размер партии 2, матрицы 2x2):

lhs = { { {1.0, 2.0},
{3.0, 4.0} },
{ {5.0, 6.0},
{7.0, 8.0} } }

rhs = { { {1.0, 0.0},
{0.0, 1.0} },
{ {1.0, 0.0},
{0.0, 1.0} } }

DotDimensionNumbers dnums;
dnums.add_lhs_contracting_dimensions(2);
dnums.add_rhs_contracting_dimensions(1);
dnums.add_lhs_batch_dimensions(0);
dnums.add_rhs_batch_dimensions(0);

DotGeneral(lhs, rhs, dnums) -> { { {1.0, 2.0},
{3.0, 4.0} },
{ {5.0, 6.0},
{7.0, 8.0} } }
Вход Вывод Семантика
[b0, m, k] dot [b0, k, n] [б0, м, н] пакетный матмуль
[b0, b1, m, k] dot [b0, b1, k, n] [b0, b1, м, н] пакетный матмуль

Из этого следует, что результирующий размерный номер начинается с размерности партии, затем с размерности «левый» без контракта/непакетный размер и, наконец, с размерности «правый» без контракта/непакетный.

Динамический срез

См. также XlaBuilder::DynamicSlice .

DynamicSlice извлекает подмассив из входного массива в динамическом start_indices . Размер среза в каждом измерении передается в size_indices , которые определяют конечную точку исключительных интервалов среза в каждом измерении: [начало, начало + размер). Форма start_indices должна иметь ранг == 1, а размер измерения равен рангу operand .

DynamicSlice(operand, start_indices, size_indices)

Аргументы Тип Семантика
operand XlaOp N-мерный массив типа T
start_indices последовательность N XlaOp Список из N скалярных целых чисел, содержащих начальные индексы среза для каждого измерения. Value must be greater than or equal to zero.
size_indices ArraySlice<int64> List of N integers containing the slice size for each dimension. Each value must be strictly greater than zero, and start + size must be less than or equal to the size of the dimension to avoid wrapping modulo dimension size.

The effective slice indices are computed by applying the following transformation for each index i in [1, N) before performing the slice:

start_indices[i] = clamp(start_indices[i], 0, operand.dimension_size[i] - size_indices[i])

This ensures that the extracted slice is always in-bounds with respect to the operand array. If the slice is in-bounds before the transformation is applied, the transformation has no effect.

1-dimensional example:

let a = {0.0, 1.0, 2.0, 3.0, 4.0}
let s = {2}

DynamicSlice(a, s, {2}) produces:
{2.0, 3.0}

2-dimensional example:

let b =
{ {0.0,  1.0,  2.0},
{3.0,  4.0,  5.0},
{6.0,  7.0,  8.0},
{9.0, 10.0, 11.0} }
let s = {2, 1}

DynamicSlice(b, s, {2, 2}) produces:
{ { 7.0,  8.0},
{10.0, 11.0} }

DynamicUpdateSlice

See also XlaBuilder::DynamicUpdateSlice .

DynamicUpdateSlice generates a result which is the value of the input array operand , with a slice update overwritten at start_indices . The shape of update determines the shape of the sub-array of the result which is updated. The shape of start_indices must be rank == 1, with dimension size equal to the rank of operand .

DynamicUpdateSlice(operand, update, start_indices)

Arguments Type Semantics
operand XlaOp N dimensional array of type T
update XlaOp N dimensional array of type T containing the slice update. Each dimension of update shape must be strictly greater than zero, and start + update must be less than or equal to the operand size for each dimension to avoid generating out-of-bounds update indices.
start_indices sequence of N XlaOp List of N scalar integers containing the starting indices of the slice for each dimension. Value must be greater than or equal to zero.

The effective slice indices are computed by applying the following transformation for each index i in [1, N) before performing the slice:

start_indices[i] = clamp(start_indices[i], 0, operand.dimension_size[i] - update.dimension_size[i])

This ensures that the updated slice is always in-bounds with respect to the operand array. If the slice is in-bounds before the transformation is applied, the transformation has no effect.

1-dimensional example:

let a = {0.0, 1.0, 2.0, 3.0, 4.0}
let u = {5.0, 6.0}
let s = {2}

DynamicUpdateSlice(a, u, s) produces:
{0.0, 1.0, 5.0, 6.0, 4.0}

2-dimensional example:

let b =
{ {0.0,  1.0,  2.0},
{3.0,  4.0,  5.0},
{6.0,  7.0,  8.0},
{9.0, 10.0, 11.0} }
let u =
{ {12.0,  13.0},
{14.0,  15.0},
{16.0,  17.0} }

let s = {1, 1}

DynamicUpdateSlice(b, u, s) produces:
{ {0.0,  1.0,  2.0},
{3.0, 12.0, 13.0},
{6.0, 14.0, 15.0},
{9.0, 16.0, 17.0} }

Element-wise binary arithmetic operations

See also XlaBuilder::Add .

A set of element-wise binary arithmetic operations is supported.

Op(lhs, rhs)

Where Op is one of Add (addition), Sub (subtraction), Mul (multiplication), Div (division), Rem (remainder), Max (maximum), Min (minimum), LogicalAnd (logical AND), or LogicalOr (logical OR).

Arguments Type Semantics
lhs XlaOp left-hand-side operand: array of type T
rhs XlaOp right-hand-side operand: array of type T

The arguments' shapes have to be either similar or compatible. See the broadcasting documentation about what it means for shapes to be compatible. The result of an operation has a shape which is the result of broadcasting the two input arrays. In this variant, operations between arrays of different ranks are not supported, unless one of the operands is a scalar.

When Op is Rem , the sign of the result is taken from the dividend, and the absolute value of the result is always less than the divisor's absolute value.

Integer division overflow (signed/unsigned division/remainder by zero or signed division/remainder of INT_SMIN with -1 ) produces an implementation defined value.

An alternative variant with different-rank broadcasting support exists for these operations:

Op(lhs, rhs, broadcast_dimensions)

Where Op is the same as above. This variant of the operation should be used for arithmetic operations between arrays of different ranks (such as adding a matrix to a vector).

The additional broadcast_dimensions operand is a slice of integers used to expand the rank of the lower-rank operand up to the rank of the higher-rank operand. broadcast_dimensions maps the dimensions of the lower-rank shape to the dimensions of the higher-rank shape. The unmapped dimensions of the expanded shape are filled with dimensions of size one. Degenerate-dimension broadcasting then broadcasts the shapes along these degenerate dimensions to equalize the shapes of both operands. The semantics are described in detail on the broadcasting page .

Element-wise comparison operations

See also XlaBuilder::Eq .

A set of standard element-wise binary comparison operations is supported. Note that standard IEEE 754 floating-point comparison semantics apply when comparing floating-point types.

Op(lhs, rhs)

Where Op is one of Eq (equal-to), Ne (not equal-to), Ge (greater-or-equal-than), Gt (greater-than), Le (less-or-equal-than), Lt (less-than). Another set of operators, EqTotalOrder, NeTotalOrder, GeTotalOrder, GtTotalOrder, LeTotalOrder, and LtTotalOrder, provide the same functionalities, except that they additionally support a total order over the floating point numbers, by enforcing -NaN < -Inf < -Finite < -0 < +0 < +Finite < +Inf < +NaN.

Arguments Type Semantics
lhs XlaOp left-hand-side operand: array of type T
rhs XlaOp right-hand-side operand: array of type T

The arguments' shapes have to be either similar or compatible. See the broadcasting documentation about what it means for shapes to be compatible. The result of an operation has a shape which is the result of broadcasting the two input arrays with the element type PRED . In this variant, operations between arrays of different ranks are not supported, unless one of the operands is a scalar.

An alternative variant with different-rank broadcasting support exists for these operations:

Op(lhs, rhs, broadcast_dimensions)

Where Op is the same as above. This variant of the operation should be used for comparison operations between arrays of different ranks (such as adding a matrix to a vector).

The additional broadcast_dimensions operand is a slice of integers specifying the dimensions to use for broadcasting the operands. The semantics are described in detail on the broadcasting page .

Element-wise unary functions

XlaBuilder supports these element-wise unary functions:

Abs(operand) Element-wise abs x -> |x| .

Ceil(operand) Element-wise ceil x -> ⌈x⌉ .

Cos(operand) Element-wise cosine x -> cos(x) .

Exp(operand) Element-wise natural exponential x -> e^x .

Floor(operand) Element-wise floor x -> ⌊x⌋ .

Imag(operand) Element-wise imaginary part of a complex (or real) shape. x -> imag(x) . If the operand is a floating point type, returns 0.

IsFinite(operand) Tests whether each element of operand is finite, ie, is not positive or negative infinity, and is not NaN . Returns an array of PRED values with the same shape as the input, where each element is true if and only if the corresponding input element is finite.

Log(operand) Element-wise natural logarithm x -> ln(x) .

LogicalNot(operand) Element-wise logical not x -> !(x) .

Logistic(operand) Element-wise logistic function computation x -> logistic(x) .

PopulationCount(operand) Computes the number of bits set in each element of operand .

Neg(operand) Element-wise negation x -> -x .

Real(operand) Element-wise real part of a complex (or real) shape. x -> real(x) . If the operand is a floating point type, returns the same value.

Rsqrt(operand) Element-wise reciprocal of square root operation x -> 1.0 / sqrt(x) .

Sign(operand) Element-wise sign operation x -> sgn(x) where

\[\text{sgn}(x) = \begin{cases} -1 & x < 0\\ -0 & x = -0\\ NaN & x = NaN\\ +0 & x = +0\\ 1 & x > 0 \end{cases}\]

using the comparison operator of the element type of operand .

Sqrt(operand) Element-wise square root operation x -> sqrt(x) .

Cbrt(operand) Element-wise cubic root operation x -> cbrt(x) .

Tan(operand) Element-wise tangent x -> tan(x) .

Tanh(operand) Element-wise hyperbolic tangent x -> tanh(x) .

Round(operand) Element-wise rounding, ties away from zero.

RoundNearestEven(operand) Element-wise rounding, ties to nearest even.

Arguments Type Semantics
operand XlaOp The operand to the function

The function is applied to each element in the operand array, resulting in an array with the same shape. It is allowed for operand to be a scalar (rank 0).

Fft

The XLA FFT operation implements the forward and inverse Fourier Transforms for real and complex inputs/outputs. Multidimensional FFTs on up to 3 axes are supported.

See also XlaBuilder::Fft .

Arguments Type Semantics
operand XlaOp The array we are Fourier transforming.
fft_type FftType See the table below.
fft_length ArraySlice<int64> The time-domain lengths of the axes being transformed. This is needed in particular for IRFFT to right-size the innermost axis, since RFFT(fft_length=[16]) has the same output shape as RFFT(fft_length=[17]) .
FftType Semantics
FFT Forward complex-to-complex FFT. Shape is unchanged.
IFFT Inverse complex-to-complex FFT. Shape is unchanged.
RFFT Forward real-to-complex FFT. Shape of the innermost axis is reduced to fft_length[-1] // 2 + 1 if fft_length[-1] is a non-zero value, omitting the reversed conjugate part of the transformed signal beyond the Nyquist frequency.
IRFFT Inverse real-to-complex FFT (ie takes complex, returns real). Shape of the innermost axis is expanded to fft_length[-1] if fft_length[-1] is a non-zero value, inferring the part of the transformed signal beyond the Nyquist frequency from the reverse conjugate of the 1 to fft_length[-1] // 2 + 1 entries.

Multidimensional FFT

When more than 1 fft_length is provided, this is equivalent to applying a cascade of FFT operations to each of the innermost axes. Note that for the real->complex and complex->real cases, the innermost axis transform is (effectively) performed first (RFFT; last for IRFFT), which is why the innermost axis is the one which changes size. Other axis transforms will then be complex->complex.

Implementation details

CPU FFT is backed by Eigen's TensorFFT. GPU FFT uses cuFFT.

Gather

The XLA gather operation stitches together several slices (each slice at a potentially different runtime offset) of an input array.

General Semantics

See also XlaBuilder::Gather . For a more intuitive description, see the "Informal Description" section below.

gather(operand, start_indices, offset_dims, collapsed_slice_dims, slice_sizes, start_index_map)

Arguments Type Semantics
operand XlaOp The array we're gathering from.
start_indices XlaOp Array containing the starting indices of the slices we gather.
index_vector_dim int64 The dimension in start_indices that "contains" the starting indices. See below for a detailed description.
offset_dims ArraySlice<int64> The set of dimensions in the output shape that offset into an array sliced from operand.
slice_sizes ArraySlice<int64> slice_sizes[i] is the bounds for the slice on dimension i .
collapsed_slice_dims ArraySlice<int64> The set of dimensions in each slice that are collapsed away. These dimensions must have size 1.
start_index_map ArraySlice<int64> A map that describes how to map indices in start_indices to legal indices into operand.
indices_are_sorted bool Whether the indices are guaranteed to be sorted by the caller.
unique_indices bool Whether the indices are guaranteed to be unique by the caller.

For convenience, we label dimensions in the output array not in offset_dims as batch_dims .

The output is an array of rank batch_dims.size + offset_dims.size .

The operand.rank must equal the sum of offset_dims.size and collapsed_slice_dims.size . Also, slice_sizes.size has to be equal to operand.rank .

If index_vector_dim is equal to start_indices.rank we implicitly consider start_indices to have a trailing 1 dimension (ie if start_indices was of shape [6,7] and index_vector_dim is 2 then we implicitly consider the shape of start_indices to be [6,7,1] ).

The bounds for the output array along dimension i is computed as follows:

  1. If i is present in batch_dims (ie is equal to batch_dims[k] for some k ) then we pick the corresponding dimension bounds out of start_indices.shape , skipping index_vector_dim (ie pick start_indices.shape.dims [ k ] if k < index_vector_dim and start_indices.shape.dims [ k + 1 ] otherwise).

  2. If i is present in offset_dims (ie equal to offset_dims [ k ] for some k ) then we pick the corresponding bound out of slice_sizes after accounting for collapsed_slice_dims (ie we pick adjusted_slice_sizes [ k ] where adjusted_slice_sizes is slice_sizes with the bounds at indices collapsed_slice_dims removed).

Formally, the operand index In corresponding to a given output index Out is calculated as follows:

  1. Let G = { Out [ k ] for k in batch_dims }. Use G to slice out a vector S such that S [ i ] = start_indices [Combine( G , i )] where Combine(A, b) inserts b at position index_vector_dim into A. Note that this is well defined even if G is empty -- if G is empty then S = start_indices .

  2. Create a starting index, S in , into operand using S by scattering S using start_index_map . More precisely:

    1. S in [ start_index_map [ k ]] = S [ k ] if k < start_index_map.size .

    2. S in [ _ ] = 0 otherwise.

  3. Create an index O in into operand by scattering the indices at the offset dimensions in Out according to the collapsed_slice_dims set. More precisely:

    1. O in [ remapped_offset_dims ( k )] = Out [ offset_dims [ k ]] if k < offset_dims.size ( remapped_offset_dims is defined below).

    2. O in [ _ ] = 0 otherwise.

  4. In is O in + S in where + is element-wise addition.

remapped_offset_dims is a monotonic function with domain [ 0 , offset_dims.size ) and range [ 0 , operand.rank ) \ collapsed_slice_dims . So if, eg, offset_dims.size is 4 , operand.rank is 6 and collapsed_slice_dims is { 0 , 2 } then remapped_offset_dims is { 01 , 13 , 24 , 35 }.

If indices_are_sorted is set to true then XLA can assume that start_indices are sorted (in ascending start_index_map order) by the user. If they are not then the semantics is implementation defined.

If unique_indices is set to true then XLA can assume that all element scattered to are unique. So XLA could use non-atomic operations. If unique_indices is set to true and the indices being scattered to are not unique then the semantics is implementation defined.

Informal Description and Examples

Informally, every index Out in the output array corresponds to an element E in the operand array, computed as follows:

  • We use the batch dimensions in Out to look up a starting index from start_indices .

  • We use start_index_map to map the starting index (whose size may be less than operand.rank) to a "full" starting index into the operand .

  • We dynamic-slice out a slice with size slice_sizes using the full starting index.

  • We reshape the slice by collapsing the collapsed_slice_dims dimensions. Since all collapsed slice dimensions must have a bound of 1, this reshape is always legal.

  • We use the offset dimensions in Out to index into this slice to get the input element, E , corresponding to output index Out .

index_vector_dim is set to start_indices.rank - 1 in all of the examples that follow. More interesting values for index_vector_dim do not change the operation fundamentally, but make the visual representation more cumbersome.

To get an intuition on how all of the above fits together, let's look at an example that gathers 5 slices of shape [8,6] from a [16,11] array. The position of a slice into the [16,11] array can be represented as an index vector of shape S64[2] , so the set of 5 positions can be represented as a S64[5,2] array.

The behavior of the gather operation can then be depicted as an index transformation that takes [ G , O 0 , O 1 ], an index in the output shape, and maps it to an element in the input array in the following way:

We first select an ( X , Y ) vector from the gather indices array using G . The element in the output array at index [ G , O 0 , O 1 ] is then the element in the input array at index [ X + O 0 , Y + O 1 ].

slice_sizes is [8,6] , which decides the range of O 0 and O 1 , and this in turn decides the bounds of the slice.

This gather operation acts as a batch dynamic slice with G as the batch dimension.

The gather indices may be multidimensional. For instance, a more general version of the example above using a "gather indices" array of shape [4,5,2] would translate indices like this:

Again, this acts as a batch dynamic slice G 0 and G 1 as the batch dimensions. The slice size is still [8,6] .

The gather operation in XLA generalizes the informal semantics outlined above in the following ways:

  1. We can configure which dimensions in the output shape are the offset dimensions (dimensions containing O 0 , O 1 in the last example). The output batch dimensions (dimensions containing G 0 , G 1 in the last example) are defined to be the output dimensions that are not offset dimensions.

  2. The number of output offset dimensions explicitly present in the output shape may be smaller than the input rank. These "missing" dimensions, which are listed explicitly as collapsed_slice_dims , must have a slice size of 1 . Since they have a slice size of 1 the only valid index for them is 0 and eliding them does not introduce ambiguity.

  3. The slice extracted from the "Gather Indices" array (( X , Y ) in the last example) may have fewer elements than the input array rank, and an explicit mapping dictates how the index should be expanded to have the same rank as the input.

As a final example, we use (2) and (3) to implement tf.gather_nd :

G 0 and G 1 are used to slice out a starting index from the gather indices array as usual, except the starting index has only one element, X . Similarly, there is only one output offset index with the value O 0 . However, before being used as indices into the input array, these are expanded in accordance to "Gather Index Mapping" ( start_index_map in the formal description) and "Offset Mapping" ( remapped_offset_dims in the formal description) into [ X , 0 ] and [ 0 , O 0 ] respectively, adding up to [ X , O 0 ]. In other words, the output index [ G 0 , G 1 , O 0 ] maps to the input index [ GatherIndices [ G 0 , G 1 , 0 ], O 0 ] which gives us the semantics for tf.gather_nd .

slice_sizes for this case is [1,11] . Intuitively this means that every index X in the gather indices array picks an entire row and the result is the concatenation of all these rows.

GetDimensionSize

See also XlaBuilder::GetDimensionSize .

Returns the size of the given dimension of the operand. The operand must be array shaped.

GetDimensionSize(operand, dimension)

Arguments Type Semantics
operand XlaOp n dimensional input array
dimension int64 A value in the interval [0, n) that specifies the dimension

SetDimensionSize

See also XlaBuilder::SetDimensionSize .

Sets the dynamic size of XlaOp's given dimension. The operand must be array shaped.

SetDimensionSize(operand, size, dimension)

Arguments Type Semantics
operand XlaOp n dimensional input array.
size XlaOp int32 representing the runtime dynamic size.
dimension int64 A value in the interval [0, n) that specifies the dimension.

Pass through the operand as result, with dynamic dimension tracked by the compiler.

Padded values will be ignored by downstream reduction ops.

let v: f32[10] = f32[10]{1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
let five: s32 = 5;
let six: s32 = 6;

// Setting dynamic dimension size doesn't change the upper bound of the static
// shape.
let padded_v_five: f32[10] = set_dimension_size(v, five, /*dimension=*/0);
let padded_v_six: f32[10] = set_dimension_size(v, six, /*dimension=*/0);

// sum == 1 + 2 + 3 + 4 + 5
let sum:f32[] = reduce_sum(padded_v_five);
// product == 1 * 2 * 3 * 4 * 5
let product:f32[] = reduce_product(padded_v_five);

// Changing padding size will yield different result.
// sum == 1 + 2 + 3 + 4 + 5 + 6
let sum:f32[] = reduce_sum(padded_v_six);

GetTupleElement

See also XlaBuilder::GetTupleElement .

Indexes into a tuple with a compile-time-constant value.

The value must be a compile-time-constant so that shape inference can determine the type of the resulting value.

This is analogous to std::get<int N>(t) in C++. Conceptually:

let v: f32[10] = f32[10]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
let s: s32 = 5;
let t: (f32[10], s32) = tuple(v, s);
let element_1: s32 = gettupleelement(t, 1);  // Inferred shape matches s32.

See also tf.tuple .

Infeed

See also XlaBuilder::Infeed .

Infeed(shape)

Argument Type Semantics
shape Shape Shape of the data read from the Infeed interface. The layout field of the shape must be set to match the layout of the data sent to the device; otherwise its behavior is undefined.

Reads a single data item from the implicit Infeed streaming interface of the device, interpreting the data as the given shape and its layout, and returns a XlaOp of the data. Multiple Infeed operations are allowed in a computation, but there must be a total order among the Infeed operations. For example, two Infeeds in the code below have a total order since there is a dependency between the while loops.

result1 = while (condition, init = init_value) {
  Infeed(shape)
}

result2 = while (condition, init = result1) {
  Infeed(shape)
}

Nested tuple shapes are not supported. For an empty tuple shape, the Infeed operation is effectively a no-op and proceeds without reading any data from the Infeed of the device.

Iota

See also XlaBuilder::Iota .

Iota(shape, iota_dimension)

Builds a constant literal on device rather than a potentially large host transfer. Creates an array that has specified shape and holds values starting at zero and incrementing by one along the specified dimension. For floating-point types, the produced array is equivalent to ConvertElementType(Iota(...)) where the Iota is of integral type and the conversion is to the floating-point type.

Arguments Type Semantics
shape Shape Shape of the array created by Iota()
iota_dimension int64 The dimension to increment along.

For example, Iota(s32[4, 8], 0) returns

  [[0, 0, 0, 0, 0, 0, 0, 0 ],
   [1, 1, 1, 1, 1, 1, 1, 1 ],
   [2, 2, 2, 2, 2, 2, 2, 2 ],
   [3, 3, 3, 3, 3, 3, 3, 3 ]]

Iota(s32[4, 8], 1) returns

  [[0, 1, 2, 3, 4, 5, 6, 7 ],
   [0, 1, 2, 3, 4, 5, 6, 7 ],
   [0, 1, 2, 3, 4, 5, 6, 7 ],
   [0, 1, 2, 3, 4, 5, 6, 7 ]]

Map

See also XlaBuilder::Map .

Map(operands..., computation)

Arguments Type Semantics
operands sequence of N XlaOp s N arrays of types T 0..T {N-1}
computation XlaComputation computation of type T_0, T_1, ..., T_{N + M -1} -> S with N parameters of type T and M of arbitrary type
dimensions int64 array array of map dimensions

Applies a scalar function over the given operands arrays, producing an array of the same dimensions where each element is the result of the mapped function applied to the corresponding elements in the input arrays.

The mapped function is an arbitrary computation with the restriction that it has N inputs of scalar type T and a single output with type S . The output has the same dimensions as the operands except that the element type T is replaced with S.

For example: Map(op1, op2, op3, computation, par1) maps elem_out <- computation(elem1, elem2, elem3, par1) at each (multi-dimensional) index in the input arrays to produce the output array.

OptimizationBarrier

Blocks any optimization pass from moving computations across the barrier.

Ensures that all inputs are evaluated before any operators that depend on the barrier's outputs.

Pad

See also XlaBuilder::Pad .

Pad(operand, padding_value, padding_config)

Arguments Type Semantics
operand XlaOp array of type T
padding_value XlaOp scalar of type T to fill in the added padding
padding_config PaddingConfig padding amount on both edges (low, high) and between the elements of each dimension

Expands the given operand array by padding around the array as well as between the elements of the array with the given padding_value . padding_config specifies the amount of edge padding and the interior padding for each dimension.

PaddingConfig is a repeated field of PaddingConfigDimension , which contains three fields for each dimension: edge_padding_low , edge_padding_high , and interior_padding .

edge_padding_low and edge_padding_high specify the amount of padding added at the low-end (next to index 0) and the high-end (next to the highest index) of each dimension respectively. The amount of edge padding can be negative -- the absolute value of negative padding indicates the number of elements to remove from the specified dimension.

interior_padding specifies the amount of padding added between any two elements in each dimension; it may not be negative. Interior padding occurs logically before edge padding, so in the case of negative edge padding, elements are removed from the interior-padded operand.

This operation is a no-op if the edge padding pairs are all (0, 0) and the interior padding values are all 0. The figure below shows examples of different edge_padding and interior_padding values for a two-dimensional array.

Recv

See also XlaBuilder::Recv .

Recv(shape, channel_handle)

Arguments Type Semantics
shape Shape shape of the data to receive
channel_handle ChannelHandle unique identifier for each send/recv pair

Receives data of the given shape from a Send instruction in another computation that shares the same channel handle. Returns a XlaOp for the received data.

The client API of Recv operation represents synchronous communication. However, the instruction is internally decomposed into 2 HLO instructions ( Recv and RecvDone ) to enable asynchronous data transfers. See also HloInstruction::CreateRecv and HloInstruction::CreateRecvDone .

Recv(const Shape& shape, int64 channel_id)

Allocates resources required to receive data from a Send instruction with the same channel_id. Returns a context for the allocated resources, which is used by a following RecvDone instruction to wait for the completion of the data transfer. The context is a tuple of {receive buffer (shape), request identifier (U32)} and it can only be used by a RecvDone instruction.

RecvDone(HloInstruction context)

Given a context created by a Recv instruction, waits for the data transfer to complete and returns the received data.

Reduce

See also XlaBuilder::Reduce .

Applies a reduction function to one or more arrays in parallel.

Reduce(operands..., init_values..., computation, dimensions)

Arguments Type Semantics
operands Sequence of N XlaOp N arrays of types T_0, ..., T_{N-1} .
init_values Sequence of N XlaOp N scalars of types T_0, ..., T_{N-1} .
computation XlaComputation computation of type T_0, ..., T_{N-1}, T_0, ..., T_{N-1} -> Collate(T_0, ..., T_{N-1}) .
dimensions int64 array unordered array of dimensions to reduce.

Where:

  • N is required to be greater or equal to 1.
  • The computation has to be "roughly" associative (see below).
  • All input arrays must have the same dimensions.
  • All initial values have to form an identity under computation .
  • If N = 1 , Collate(T) is T .
  • If N > 1 , Collate(T_0, ..., T_{N-1}) is a tuple of N elements of type T .

This operation reduces one or more dimensions of each input array into scalars. The rank of each returned array is rank(operand) - len(dimensions) . The output of the op is Collate(Q_0, ..., Q_N) where Q_i is an array of type T_i , the dimensions of which are described below.

Different backends are allowed to reassociate the reduction computation. This can lead to numerical differences, as some reduction functions like addition are not associative for floats. However, if the range of the data is limited, floating-point addition is close enough to being associative for most practical uses.

Examples

When reducing across one dimension in a single 1D array with values [10, 11, 12, 13] , with reduction function f (this is computation ) then that could be computed as

f(10, f(11, f(12, f(init_value, 13)))

but there are also many other possibilities, eg

f(init_value, f(f(10, f(init_value, 11)), f(f(init_value, 12), f(init_value, 13))))

The following is a rough pseudo-code example of how reduction could be implemented, using summation as the reduction computation with an initial value of 0.

result_shape <- remove all dims in dimensions from operand_shape

# Iterate over all elements in result_shape. The number of r's here is equal
# to the rank of the result
for r0 in range(result_shape[0]), r1 in range(result_shape[1]), ...:
  # Initialize this result element
  result[r0, r1...] <- 0

  # Iterate over all the reduction dimensions
  for d0 in range(dimensions[0]), d1 in range(dimensions[1]), ...:
    # Increment the result element with the value of the operand's element.
    # The index of the operand's element is constructed from all ri's and di's
    # in the right order (by construction ri's and di's together index over the
    # whole operand shape).
    result[r0, r1...] += operand[ri... di]

Here's an example of reducing a 2D array (matrix). The shape has rank 2, dimension 0 of size 2 and dimension 1 of size 3:

Results of reducing dimensions 0 or 1 with an "add" function:

Note that both reduction results are 1D arrays. The diagram shows one as column and another as row just for visual convenience.

For a more complex example, here is a 3D array. Its rank is 3, dimension 0 of size 4, dimension 1 of size 2 and dimension 2 of size 3. For simplicity, the values 1 to 6 are replicated across dimension 0.

Similarly to the 2D example, we can reduce just one dimension. If we reduce dimension 0, for example, we get a rank-2 array where all values across dimension 0 were folded into a scalar:

|  4   8  12 |
| 16  20  24 |

If we reduce dimension 2, we also get a rank-2 array where all values across dimension 2 were folded into a scalar:

| 6  15 |
| 6  15 |
| 6  15 |
| 6  15 |

Note that the relative order between the remaining dimensions in the input is preserved in the output, but some dimensions may get assigned new numbers (since the rank changes).

We can also reduce multiple dimensions. Add-reducing dimensions 0 and 1 produces the 1D array [20, 28, 36] .

Reducing the 3D array over all its dimensions produces the scalar 84 .

Variadic Reduce

When N > 1 , reduce function application is slightly more complex, as it is applied simultaneously to all inputs. The operands are supplied to the computation in the following order:

  • Running reduced value for the first operand
  • ...
  • Running reduced value for the N'th operand
  • Input value for the first operand
  • ...
  • Input value for the N'th operand

For example, consider the following reduction function, which can be used to compute the max and the argmax of a 1-D array in parallel:

f: (Float, Int, Float, Int) -> Float, Int
f(max, argmax, value, index):
  if value >= max:
    return (value, index)
  else:
    return (max, argmax)

For 1-D Input arrays V = Float[N], K = Int[N] , and init values I_V = Float, I_K = Int , the result f_(N-1) of reducing across the only input dimension is equivalent to the following recursive application:

f_0 = f(I_V, I_K, V_0, K_0)
f_1 = f(f_0.first, f_0.second, V_1, K_1)
...
f_(N-1) = f(f_(N-2).first, f_(N-2).second, V_(N-1), K_(N-1))

Applying this reduction to an array of values, and an array of sequential indices (ie iota), will co-iterate over the arrays, and return a tuple containing the maximal value and the matching index.

ReducePrecision

See also XlaBuilder::ReducePrecision .

Models the effect of converting floating-point values to a lower-precision format (such as IEEE-FP16) and back to the original format. The number of exponent and mantissa bits in the lower-precision format can be specified arbitrarily, although all bit sizes may not be supported on all hardware implementations.

ReducePrecision(operand, mantissa_bits, exponent_bits)

Arguments Type Semantics
operand XlaOp array of floating-point type T .
exponent_bits int32 number of exponent bits in lower-precision format
mantissa_bits int32 number of mantissa bits in lower-precision format

The result is an array of type T . The input values are rounded to the nearest value representable with the given number of mantissa bits (using "ties to even" semantics), and any values that exceed the range specified by the number of exponent bits are clamped to positive or negative infinity. NaN values are retained, although they may be converted to canonical NaN values.

The lower-precision format must have at least one exponent bit (in order to distinguish a zero value from an infinity, since both have a zero mantissa), and must have a non-negative number of mantissa bits. The number of exponent or mantissa bits may exceed the corresponding value for type T ; the corresponding portion of the conversion is then simply a no-op.

ReduceScatter

See also XlaBuilder::ReduceScatter .

ReduceScatter is a collective operation that effectively does an AllReduce and then scatters the result by splitting it into shard_count blocks along the scatter_dimension and replica i in the replica group receives the ith shard.

ReduceScatter(operand, computation, scatter_dim, shard_count, replica_group_ids, channel_id)

Arguments Type Semantics
operand XlaOp Array or a non-empty tuple of arrays to reduce across replicas.
computation XlaComputation Reduction computation
scatter_dimension int64 Dimension to scatter.
shard_count int64 Number of blocks to split scatter_dimension
replica_groups vector of vectors of int64 Groups between which the reductions are performed
channel_id optional int64 Optional channel ID for cross-module communication
  • When operand is a tuple of arrays, the reduce-scatter is performed on each element of the tuple.
  • replica_groups is a list of replica groups between which the reduction is performed (replica id for the current replica can be retrieved using ReplicaId ). The order of replicas in each group determines the order in which the all-reduce result will be scattered. replica_groups must either be empty (in which case all replicas belong to a single group), or contain the same number of elements as the number of replicas. When there are more than one replica groups, they all must be of the same size. For example, replica_groups = {0, 2}, {1, 3} performs reduction between the replicas 0 and 2 , and 1 and 3 and then scatters the result.
  • shard_count is the size of each replica group. We need this in cases where replica_groups are empty. If replica_groups is not empty, shard_count must be equal to the size of each replica group.
  • channel_id is used for cross-module communication: only reduce-scatter operations with the same channel_id can communicate with each other.

The output shape is the input shape with the scatter_dimension made shard_count times smaller. For example, if there are two replicas and the operand has the value [1.0, 2.25] and [3.0, 5.25] respectively on the two replicas, then the output value from this op where scatter_dim is 0 will be [4.0] for the first replica and [7.5] for the second replica.

ReduceWindow

See also XlaBuilder::ReduceWindow .

Applies a reduction function to all elements in each window of a sequence of N multi-dimensional arrays, producing a single or a tuple of N multi-dimensional arrays as output. Each output array has the same number of elements as the number of valid positions of the window. A pooling layer can be expressed as a ReduceWindow . Similar to Reduce , the applied computation is always passed the init_values on the left-hand side.

ReduceWindow(operands..., init_values..., computation, window_dimensions, window_strides, padding)

Arguments Type Semantics
operands N XlaOps A sequence of N multi-dimensional arrays of types T_0,..., T_{N-1} , each representing the base area on which the window is placed.
init_values N XlaOps The N starting values for the reduction, one for each of the N operands. See Reduce for details.
computation XlaComputation Reduction function of type T_0, ..., T_{N-1}, T_0, ..., T_{N-1} -> Collate(T_0, ..., T_{N-1}) , to apply to elements in each window of all the input operands.
window_dimensions ArraySlice<int64> array of integers for window dimension values
window_strides ArraySlice<int64> array of integers for window stride values
base_dilations ArraySlice<int64> array of integers for base dilation values
window_dilations ArraySlice<int64> array of integers for window dilation values
padding Padding padding type for window (Padding::kSame, which pads so as to have the same output shape as input if the stride is 1, or Padding::kValid, which uses no padding and "stops" the window once it no longer fits)

Where:

  • N is required to be greater or equal to 1.
  • All input arrays must have the same dimensions.
  • If N = 1 , Collate(T) is T .
  • If N > 1 , Collate(T_0, ..., T_{N-1}) is a tuple of N elements of type (T0,...T{N-1}) .

Below code and figure shows an example of using ReduceWindow . Input is a matrix of size [4x6] and both window_dimensions and window_stride_dimensions are [2x3].

// Create a computation for the reduction (maximum).
XlaComputation max;
{
  XlaBuilder builder(client_, "max");
  auto y = builder.Parameter(0, ShapeUtil::MakeShape(F32, {}), "y");
  auto x = builder.Parameter(1, ShapeUtil::MakeShape(F32, {}), "x");
  builder.Max(y, x);
  max = builder.Build().value();
}

// Create a ReduceWindow computation with the max reduction computation.
XlaBuilder builder(client_, "reduce_window_2x3");
auto shape = ShapeUtil::MakeShape(F32, {4, 6});
auto input = builder.Parameter(0, shape, "input");
builder.ReduceWindow(
    input,
    /*init_val=*/builder.ConstantLiteral(LiteralUtil::MinValue(F32)),
    *max,
    /*window_dimensions=*/{2, 3},
    /*window_stride_dimensions=*/{2, 3},
    Padding::kValid);

Stride of 1 in a dimension specifies that the position of a window in the dimension is 1 element away from its adjacent window. In order to specify that no windows overlap with each other, window_stride_dimensions should be equal to window_dimensions. The figure below illustrates the use of two different stride values. Padding is applied to each dimension of the input and the calculations are the same as though the input came in with the dimensions it has after padding.

For a non-trivial padding example, consider computing reduce-window minimum (initial value is MAX_FLOAT ) with dimension 3 and stride 2 over the input array [10000, 1000, 100, 10, 1] . Padding kValid computes minimums over two valid windows: [10000, 1000, 100] and [100, 10, 1] , resulting in the output [100, 1] . Padding kSame first pads the array so that the shape after the reduce-window would be the same as input for stride one by adding initial elements on both sides, getting [MAX_VALUE, 10000, 1000, 100, 10, 1, MAX_VALUE] . Running reduce-window over the padded array operates on three windows [MAX_VALUE, 10000, 1000] , [1000, 100, 10] , [10, 1, MAX_VALUE] , and yields [1000, 10, 1] .

The evaluation order of the reduction function is arbitrary and may be non-deterministic. Therefore, the reduction function should not be overly sensitive to reassociation. See the discussion about associativity in the context of Reduce for more details.

ReplicaId

See also XlaBuilder::ReplicaId .

Returns the unique ID (U32 scalar) of the replica.

ReplicaId()

The unique ID of each replica is an unsigned integer in the interval [0, N) , where N is the number of replicas. Since all the replicas are running the same program, a ReplicaId() call in the program will return a different value on each replica.

Reshape

See also XlaBuilder::Reshape and the Collapse operation.

Reshapes the dimensions of an array into a new configuration.

Reshape(operand, new_sizes) Reshape(operand, dimensions, new_sizes)

Arguments Type Semantics
operand XlaOp array of type T
dimensions int64 vector order in which dimensions are collapsed
new_sizes int64 vector vector of sizes of new dimensions

Conceptually, reshape first flattens an array into a one-dimensional vector of data values, and then refines this vector into a new shape. The input arguments are an arbitrary array of type T, a compile-time-constant vector of dimension indices, and a compile-time-constant vector of dimension sizes for the result. The values in the dimension vector, if given, must be a permutation of all of T's dimensions; the default if not given is {0, ..., rank - 1} . The order of the dimensions in dimensions is from slowest-varying dimension (most major) to fastest-varying dimension (most minor) in the loop nest which collapses the input array into a single dimension. The new_sizes vector determines the size of the output array. The value at index 0 in new_sizes is the size of dimension 0, the value at index 1 is the size of dimension 1, and so on. The product of the new_size dimensions must equal the product of the operand's dimension sizes. When refining the collapsed array into the multidimensional array defined by new_sizes , the dimensions in new_sizes are ordered from slowest varying (most major) and to fastest varying (most minor).

For example, let v be an array of 24 elements:

let v = f32[4x2x3] { { {10, 11, 12}, {15, 16, 17} },
                    { {20, 21, 22}, {25, 26, 27} },
                    { {30, 31, 32}, {35, 36, 37} },
                    { {40, 41, 42}, {45, 46, 47} } };

In-order collapse:
let v012_24 = Reshape(v, {0,1,2}, {24});
then v012_24 == f32[24] {10, 11, 12, 15, 16, 17, 20, 21, 22, 25, 26, 27,
                         30, 31, 32, 35, 36, 37, 40, 41, 42, 45, 46, 47};

let v012_83 = Reshape(v, {0,1,2}, {8,3});
then v012_83 == f32[8x3] { {10, 11, 12}, {15, 16, 17},
                          {20, 21, 22}, {25, 26, 27},
                          {30, 31, 32}, {35, 36, 37},
                          {40, 41, 42}, {45, 46, 47} };

Out-of-order collapse:
let v021_24 = Reshape(v, {1,2,0}, {24});
then v012_24 == f32[24]  {10, 20, 30, 40, 11, 21, 31, 41, 12, 22, 32, 42,
                          15, 25, 35, 45, 16, 26, 36, 46, 17, 27, 37, 47};

let v021_83 = Reshape(v, {1,2,0}, {8,3});
then v021_83 == f32[8x3] { {10, 20, 30}, {40, 11, 21},
                          {31, 41, 12}, {22, 32, 42},
                          {15, 25, 35}, {45, 16, 26},
                          {36, 46, 17}, {27, 37, 47} };


let v021_262 = Reshape(v, {1,2,0}, {2,6,2});
then v021_262 == f32[2x6x2] { { {10, 20}, {30, 40},
                              {11, 21}, {31, 41},
                              {12, 22}, {32, 42} },
                             { {15, 25}, {35, 45},
                              {16, 26}, {36, 46},
                              {17, 27}, {37, 47} } };

As a special case, reshape can transform a single-element array to a scalar and vice versa. For example,

Reshape(f32[1x1] { {5} }, {0,1}, {}) == 5;
Reshape(5, {}, {1,1}) == f32[1x1] { {5} };

Rev (reverse)

See also XlaBuilder::Rev .

Rev(operand, dimensions)

Arguments Type Semantics
operand XlaOp array of type T
dimensions ArraySlice<int64> dimensions to reverse

Reverses the order of elements in the operand array along the specified dimensions , generating an output array of the same shape. Each element of the operand array at a multidimensional index is stored into the output array at a transformed index. The multidimensional index is transformed by reversing the index in each dimension to be reversed (ie, if a dimension of size N is one of the reversing dimensions, its index i is transformed into N - 1 - i).

One use for the Rev operation is to reverse the convolution weight array along the two window dimensions during the gradient computation in neural networks.

RngNormal

See also XlaBuilder::RngNormal .

Constructs an output of a given shape with random numbers generated following the \(N(\mu, \sigma)\) normal distribution. The parameters \(\mu\) and\(\sigma\), and output shape have to have a floating point elemental type. The parameters furthermore have to be scalar valued.

RngNormal(mu, sigma, shape)

Arguments Type Semantics
mu XlaOp Scalar of type T specifying mean of generated numbers
sigma XlaOp Scalar of type T specifying standard deviation of generated numbers
shape Shape Output shape of type T

RngUniform

See also XlaBuilder::RngUniform .

Constructs an output of a given shape with random numbers generated following the uniform distribution over the interval \([a,b)\). The parameters and output element type have to be a boolean type, an integral type or a floating point types, and the types have to be consistent. The CPU and GPU backends currently only support F64, F32, F16, BF16, S64, U64, S32 and U32. Furthermore, the parameters need to be scalar valued. If \(b <= a\) the result is implementation-defined.

RngUniform(a, b, shape)

Arguments Type Semantics
a XlaOp Scalar of type T specifying lower limit of interval
b XlaOp Scalar of type T specifying upper limit of interval
shape Shape Output shape of type T

RngBitGenerator

Generates an output with a given shape filled with uniform random bits using the specified algorithm (or backend default) and returns an updated state (with the same shape as initial state) and the generated random data.

Initial state is the initial state of the current random number generation. It and the required shape and valid values are dependent on the algorithm used.

The output is guaranteed to be a deterministic function of the initial state but it is not guaranteed to be deterministic between backends and different compiler versions.

RngBitGenerator(algorithm, key, shape)

Arguments Type Semantics
algorithm RandomAlgorithm PRNG algorithm to be used.
initial_state XlaOp Initial state for the PRNG algorithm.
shape Shape Output shape for generated data.

Available values for algorithm :

Scatter

The XLA scatter operation generates a sequence of results which are the values of the input array operands , with several slices (at indices specified by scatter_indices ) updated with the sequence of values in updates using update_computation .

See also XlaBuilder::Scatter .

scatter(operands..., scatter_indices, updates..., update_computation, index_vector_dim, update_window_dims, inserted_window_dims, scatter_dims_to_operand_dims)

Arguments Type Semantics
operands Sequence of N XlaOp N arrays of types T_0, ..., T_N to be scattered into.
scatter_indices XlaOp Array containing the starting indices of the slices that must be scattered to.
updates Sequence of N XlaOp N arrays of types T_0, ..., T_N . updates[i] contains the values that must be used for scattering operands[i] .
update_computation XlaComputation Computation to be used for combining the existing values in the input array and the updates during scatter. This computation should be of type T_0, ..., T_N, T_0, ..., T_N -> Collate(T_0, ..., T_N) .
index_vector_dim int64 The dimension in scatter_indices that contains the starting indices.
update_window_dims ArraySlice<int64> The set of dimensions in updates shape that are window dimensions .
inserted_window_dims ArraySlice<int64> The set of window dimensions that must be inserted into updates shape.
scatter_dims_to_operand_dims ArraySlice<int64> A dimensions map from the scatter indices to the operand index space. This array is interpreted as mapping i to scatter_dims_to_operand_dims[i] . It has to be one-to-one and total.
indices_are_sorted bool Whether the indices are guaranteed to be sorted by the caller.

Where:

  • N is required to be greater or equal to 1.
  • operands [ 0 ], ..., operands [ N-1 ] must all have the same dimensions.
  • updates [ 0 ], ..., updates [ N-1 ] must all have the same dimensions.
  • If N = 1 , Collate(T) is T .
  • If N > 1 , Collate(T_0, ..., T_N) is a tuple of N elements of type T .

If index_vector_dim is equal to scatter_indices.rank we implicitly consider scatter_indices to have a trailing 1 dimension.

We define update_scatter_dims of type ArraySlice<int64> as the set of dimensions in updates shape that are not in update_window_dims , in ascending order.

The arguments of scatter should follow these constraints:

  • Each updates array must be of rank update_window_dims.size + scatter_indices.rank - 1 .

  • Bounds of dimension i in each updates array must conform to the following:

    • If i is present in update_window_dims (ie equal to update_window_dims [ k ] for some k ), then the bound of dimension i in updates must not exceed the corresponding bound of operand after accounting for the inserted_window_dims (ie adjusted_window_bounds [ k ], where adjusted_window_bounds contains the bounds of operand with the bounds at indices inserted_window_dims removed).
    • If i is present in update_scatter_dims (ie equal to update_scatter_dims [ k ] for some k ), then the bound of dimension i in updates must be equal to the corresponding bound of scatter_indices , skipping index_vector_dim (ie scatter_indices.shape.dims [ k ], if k < index_vector_dim and scatter_indices.shape.dims [ k+1 ] otherwise).
  • update_window_dims must be in ascending order, not have any repeating dimension numbers, and be in the range [0, updates.rank) .

  • inserted_window_dims must be in ascending order, not have any repeating dimension numbers, and be in the range [0, operand.rank) .

  • operand.rank must equal the sum of update_window_dims.size and inserted_window_dims.size .

  • scatter_dims_to_operand_dims.size must be equal to scatter_indices.shape.dims [ index_vector_dim ], and its values must be in the range [0, operand.rank) .

For a given index U in each updates array, the corresponding index I in the corresponding operands array into which this update has to be applied is computed as follows:

  1. Let G = { U [ k ] for k in update_scatter_dims }. Use G to look up an index vector S in the scatter_indices array such that S [ i ] = scatter_indices [Combine( G , i )] where Combine(A, b) inserts b at positions index_vector_dim into A.
  2. Create an index S in into operand using S by scattering S using the scatter_dims_to_operand_dims map. More formally:
    1. S in [ scatter_dims_to_operand_dims [ k ]] = S [ k ] if k < scatter_dims_to_operand_dims.size .
    2. S in [ _ ] = 0 otherwise.
  3. Create an index W in into each operands array by scattering the indices at update_window_dims in U according to inserted_window_dims . More formally:
    1. W in [ window_dims_to_operand_dims ( k )] = U [ k ] if k is in update_window_dims , where window_dims_to_operand_dims is the monotonic function with domain [ 0 , update_window_dims.size ) and range [ 0 , operand.rank ) \ inserted_window_dims . (For example, if update_window_dims.size is 4 , operand.rank is 6 , and inserted_window_dims is { 0 , 2 } then window_dims_to_operand_dims is { 01 , 13 , 24 , 35 }).
    2. W in [ _ ] = 0 otherwise.
  4. I is W in + S in where + is element-wise addition.

In summary, the scatter operation can be defined as follows.

  • Initialize output with operands , ie for all indices J , for all indices O in the operands [ J ] array:
    output [ J ][ O ] = operands [ J ][ O ]
  • For every index U in the updates [ J ] array and the corresponding index O in the operand [ J ] array, if O is a valid index for output :
    (output [ 0 ][ O ], ..., output [ N-1 ][ O ]) = update_computation ( output [ 0 ][ O ], ..., , output [ N-1 ][ O ], updates [ 0 ][ U ], ..., updates [ N-1 ][ U ])

The order in which updates are applied is non-deterministic. So, when multiple indices in updates refer to the same index in operands , the corresponding value in output will be non-deterministic.

Note that the first parameter that is passed into the update_computation will always be the current value from the output array and the second parameter will always be the value from the updates array. This is important specifically for cases when the update_computation is not commutative .

If indices_are_sorted is set to true then XLA can assume that start_indices are sorted (in ascending start_index_map order) by the user. If they are not then the semantics is implementation defined.

Informally, the scatter op can be viewed as an inverse of the gather op, ie the scatter op updates the elements in the input that are extracted by the corresponding gather op.

For a detailed informal description and examples, refer to the "Informal Description" section under Gather .

Select

See also XlaBuilder::Select .

Constructs an output array from elements of two input arrays, based on the values of a predicate array.

Select(pred, on_true, on_false)

Arguments Type Semantics
pred XlaOp array of type PRED
on_true XlaOp array of type T
on_false XlaOp array of type T

The arrays on_true and on_false must have the same shape. This is also the shape of the output array. The array pred must have the same dimensionality as on_true and on_false , with the PRED element type.

For each element P of pred , the corresponding element of the output array is taken from on_true if the value of P is true , and from on_false if the value of P is false . As a restricted form of broadcasting , pred can be a scalar of type PRED . In this case, the output array is taken wholly from on_true if pred is true , and from on_false if pred is false .

Example with non-scalar pred :

let pred: PRED[4] = {true, false, false, true};
let v1: s32[4] = {1, 2, 3, 4};
let v2: s32[4] = {100, 200, 300, 400};
==>
Select(pred, v1, v2) = s32[4]{1, 200, 300, 4};

Example with scalar pred :

let pred: PRED = true;
let v1: s32[4] = {1, 2, 3, 4};
let v2: s32[4] = {100, 200, 300, 400};
==>
Select(pred, v1, v2) = s32[4]{1, 2, 3, 4};

Selections between tuples are supported. Tuples are considered to be scalar types for this purpose. If on_true and on_false are tuples (which must have the same shape!) then pred has to be a scalar of type PRED .

SelectAndScatter

See also XlaBuilder::SelectAndScatter .

This operation can be considered as a composite operation that first computes ReduceWindow on the operand array to select an element from each window, and then scatters the source array to the indices of the selected elements to construct an output array with the same shape as the operand array. The binary select function is used to select an element from each window by applying it across each window, and it is called with the property that the first parameter's index vector is lexicographically less than the second parameter's index vector. The select function returns true if the first parameter is selected and returns false if the second parameter is selected, and the function must hold transitivity (ie, if select(a, b) and select(b, c) are true , then select(a, c) is also true ) so that the selected element does not depend on the order of the elements traversed for a given window.

The function scatter is applied at each selected index in the output array. It takes two scalar parameters:

  1. Current value at the selected index in the output array
  2. The scatter value from source that applies to the selected index

It combines the two parameters and returns a scalar value that's used to update the value at the selected index in the output array. Initially, all indices of the output array are set to init_value .

The output array has the same shape as the operand array and the source array must have the same shape as the result of applying a ReduceWindow operation on the operand array. SelectAndScatter can be used to backpropagate the gradient values for a pooling layer in a neural network.

SelectAndScatter(operand, select, window_dimensions, window_strides, padding, source, init_value, scatter)

Arguments Type Semantics
operand XlaOp array of type T over which the windows slide
select XlaComputation binary computation of type T, T -> PRED , to apply to all elements in each window; returns true if the first parameter is selected and returns false if the second parameter is selected
window_dimensions ArraySlice<int64> array of integers for window dimension values
window_strides ArraySlice<int64> array of integers for window stride values
padding Padding padding type for window (Padding::kSame or Padding::kValid)
source XlaOp array of type T with the values to scatter
init_value XlaOp scalar value of type T for the initial value of the output array
scatter XlaComputation binary computation of type T, T -> T , to apply each scatter source element with its destination element

The figure below shows examples of using SelectAndScatter , with the select function computing the maximal value among its parameters. Note that when the windows overlap, as in the figure (2) below, an index of the operand array may be selected multiple times by different windows. In the figure, the element of value 9 is selected by both of the top windows (blue and red) and the binary addition scatter function produces the output element of value 8 (2 + 6).

The evaluation order of the scatter function is arbitrary and may be non-deterministic. Therefore, the scatter function should not be overly sensitive to reassociation. See the discussion about associativity in the context of Reduce for more details.

Send

See also XlaBuilder::Send .

Send(operand, channel_handle)

Arguments Type Semantics
operand XlaOp data to send (array of type T)
channel_handle ChannelHandle unique identifier for each send/recv pair

Sends the given operand data to a Recv instruction in another computation that shares the same channel handle. Does not return any data.

Similar to the Recv operation, the client API of Send operation represents synchronous communication, and is internally decomposed into 2 HLO instructions ( Send and SendDone ) to enable asynchronous data transfers. See also HloInstruction::CreateSend and HloInstruction::CreateSendDone .

Send(HloInstruction operand, int64 channel_id)

Initiates an asynchronous transfer of the operand to the resources allocated by the Recv instruction with the same channel id. Returns a context, which is used by a following SendDone instruction to wait for the completion of the data transfer. The context is a tuple of {operand (shape), request identifier (U32)} and it can only be used by a SendDone instruction.

SendDone(HloInstruction context)

Given a context created by a Send instruction, waits for the data transfer to complete. The instruction does not return any data.

Scheduling of channel instructions

The execution order of the 4 instructions for each channel ( Recv , RecvDone , Send , SendDone ) is as below.

  • Recv happens before Send
  • Send happens before RecvDone
  • Recv happens before RecvDone
  • Send happens before SendDone

When the backend compilers generate a linear schedule for each computation that communicates via channel instructions, there must not be cycles across the computations. For example, below schedules lead to deadlocks.

Slice

See also XlaBuilder::Slice .

Slicing extracts a sub-array from the input array. The sub-array is of the same rank as the input and contains the values inside a bounding box within the input array where the dimensions and indices of the bounding box are given as arguments to the slice operation.

Slice(operand, start_indices, limit_indices, strides)

Arguments Type Semantics
operand XlaOp N dimensional array of type T
start_indices ArraySlice<int64> List of N integers containing the starting indices of the slice for each dimension. Values must be greater than or equal to zero.
limit_indices ArraySlice<int64> List of N integers containing the ending indices (exclusive) for the slice for each dimension. Each value must be greater than or equal to the respective start_indices value for the dimension and less than or equal to the size of the dimension.
strides ArraySlice<int64> List of N integers that decides the input stride of the slice. The slice picks every strides[d] element in dimension d .

1-dimensional example:

let a = {0.0, 1.0, 2.0, 3.0, 4.0}
Slice(a, {2}, {4}) produces:
  {2.0, 3.0}

2-dimensional example:

let b =
 { {0.0,  1.0,  2.0},
   {3.0,  4.0,  5.0},
   {6.0,  7.0,  8.0},
   {9.0, 10.0, 11.0} }

Slice(b, {2, 1}, {4, 3}) produces:
  { { 7.0,  8.0},
    {10.0, 11.0} }

Sort

See also XlaBuilder::Sort .

Sort(operands, comparator, dimension, is_stable)

Arguments Type Semantics
operands ArraySlice<XlaOp> The operands to sort.
comparator XlaComputation The comparator computation to use.
dimension int64 The dimension along which to sort.
is_stable bool Whether stable sorting should be used.

If only one operand is provided:

  • If the operand is a rank-1 tensor (an array), the result is a sorted array. If you want to sort the array into ascending order, the comparator should perform a less-than comparison. Formally, after the array is sorted, it holds for all index positions i, j with i < j that either comparator(value[i], value[j]) = comparator(value[j], value[i]) = false or comparator(value[i], value[j]) = true .

  • If the operand has higher rank, the operand is sorted along the provided dimension. For example, for a rank-2 tensor (a matrix), a dimension value of 0 will independently sort every column, and a dimension value of 1 will independently sort each row. If no dimension number is provided, then the last dimension is chosen by default. For the dimension which is sorted, the same sorting order applies as in the rank-1 case.

If n > 1 operands are provided:

  • All n operands must be tensors with the same dimensions. The element types of the tensors may be different.

  • All operands are sorted together, not individually. Conceptually the operands are treated as a tuple. When checking whether the elements of each operand at index positions i and j need to be swapped, the comparator is called with 2 * n scalar parameters, where parameter 2 * k corresponds to the value at position i from the k-th operand, and parameter 2 * k + 1 corresponds to the value at position j from the k-th operand. Usually, the comparator would thus compare parameters 2 * k and 2 * k + 1 with each other and possibly use other parameter pairs as tie breakers.

  • The result is a tuple that consists of the operands in sorted order (along the provided dimension, as above). The i-th operand of the tuple corresponds to the i-th operand of Sort.

For example, if there are three operands operand0 = [3, 1] , operand1 = [42, 50] , operand2 = [-3.0, 1.1] , and the comparator compares only the values of operand0 with less-than, then the output of the sort is the tuple ([1, 3], [50, 42], [1.1, -3.0]) .

If is_stable is set to true, the sort is guaranteed to be stable, that is, if there are elements which are considered to be equal by the comparator, the relative order of the equal values is preserved. Two elements e1 and e2 are equal if and only if comparator(e1, e2) = comparator(e2, e1) = false . By default, is_stable is set to false.

Transpose

See also the tf.reshape operation.

Transpose(operand)

Arguments Type Semantics
operand XlaOp The operand to transpose.
permutation ArraySlice<int64> How to permute the dimensions.

Permutes the operand dimensions with the given permutation, so ∀ i . 0 ≤ i < rank ⇒ input_dimensions[permutation[i]] = output_dimensions[i] .

This is the same as Reshape(operand, permutation, Permute(permutation, operand.shape.dimensions)).

TriangularSolve

See also XlaBuilder::TriangularSolve .

Solves systems of linear equations with lower or upper triangular coefficient matrices by forward- or back-substitution. Broadcasting along leading dimensions, this routine solves one of the matrix systems op(a) * x = b , or x * op(a) = b , for the variable x , given a and b , where op(a) is either op(a) = a , or op(a) = Transpose(a) , or op(a) = Conj(Transpose(a)) .

TriangularSolve(a, b, left_side, lower, unit_diagonal, transpose_a)

Arguments Type Semantics
a XlaOp a rank > 2 array of a complex or floating-point type with shape [..., M, M] .
b XlaOp a rank > 2 array of the same type with shape [..., M, K] if left_side is true, [..., K, M] otherwise.
left_side bool indicates whether to solve a system of the form op(a) * x = b ( true ) or x * op(a) = b ( false ).
lower bool whether to use the upper or lower triangle of a .
unit_diagonal bool if true , the diagonal elements of a are assumed to be 1 and not accessed.
transpose_a Transpose whether to use a as is, transpose it or take its conjugate transpose.

Input data is read only from the lower/upper triangle of a , depending on the value of lower . Values from the other triangle are ignored. Output data is returned in the same triangle; the values in the other triangle are implementation-defined and may be anything.

If the rank of a and b are greater than 2, they are treated as batches of matrices, where all except the minor 2 dimensions are batch dimensions. a and b must have equal batch dimensions.

Tuple

See also XlaBuilder::Tuple .

A tuple containing a variable number of data handles, each of which has its own shape.

This is analogous to std::tuple in C++. Conceptually:

let v: f32[10] = f32[10]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
let s: s32 = 5;
let t: (f32[10], s32) = tuple(v, s);

Tuples can be deconstructed (accessed) via the GetTupleElement operation.

While

See also XlaBuilder::While .

While(condition, body, init)

Arguments Type Semantics
condition XlaComputation XlaComputation of type T -> PRED which defines the termination condition of the loop.
body XlaComputation XlaComputation of type T -> T which defines the body of the loop.
init T Initial value for the parameter of condition and body .

Sequentially executes the body until the condition fails. This is similar to a typical while loop in many other languages except for the differences and restrictions listed below.

  • A While node returns a value of type T , which is the result from the last execution of the body .
  • The shape of the type T is statically determined and must be the same across all iterations.

The T parameters of the computations are initialized with the init value in the first iteration and are automatically updated to the new result from body in each subsequent iteration.

One main use case of the While node is to implement the repeated execution of training in neural networks. Simplified pseudocode is shown below with a graph that represents the computation. The code can be found in while_test.cc . The type T in this example is a Tuple consisting of an int32 for the iteration count and a vector[10] for the accumulator. For 1000 iterations, the loop keeps adding a constant vector to the accumulator.

// Pseudocode for the computation.
init = {0, zero_vector[10]} // Tuple of int32 and float[10].
result = init;
while (result(0) < 1000) {
  iteration = result(0) + 1;
  new_vector = result(1) + constant_vector[10];
  result = {iteration, new_vector};
}