The argument is a function that returns the learning rate given the current step (i.e., curren running iteration).

The argument function returns the momentum value given the current running step (i.e., iterations/mini-batches).

Sets the test part result reporter for the current thread.

A wrapper method to spawn a thread to execute Train() method.

Reduce argument for trigonometric functions. \param abs half-precision floating-point value \param k value to take quarter period \return \a abs reduc

Platform-independent arithmetic right shift. \param arg integer value in two's complement \param i shift amount (at most 31) \return \a arg right shif

Fixed point arc tangent. This uses the CORDIC algorithm in vectoring mode. \param my y coordinate as Q0.30 \param mx x coordinate as Q0.30 \param n nu

Insert the new_item to rank_list, then get the rank of it. :param rank_list: :param new_item: :return:

Nearest integer not less than half value. See also:** Documentation for [std::ceil](https://en.cppreference.com/w/cpp/numeric/math/ceil). \param arg h

Compile and initialize the model This function will automatically derive the shape of parameters in each sublayer based on the shape

Converts a sequence of [tokens|ids] using the vocab.

shallow copy calls copy constructor of singa::Tensor Returns: new tensor copied

Args: model: use_test_acc: epoch_num: how many epoch, set by scheduler train_loader:

Performs all reduce of the tensors after fusing and sparsification. Args: tensor(List of Tensors): a list of tensors to be al

generate shape multipliers for e.g. tensor of shape (3,3), stride (1,3) will have shape multipliers of (3,1) for e.g. tensor of shape (3,3), stride (3

generate a traversal_info vector based on the tensor's shape for the traverse_next function to work

Gets the user supplied key.

Multiply by power of two. This function is exact to rounding for all rounding modes. See also:** Documentation for [std::ldexp](https://en.cppreferen

Do matrix-matrix or matrix-vector multiplication. This function returns C = alpha * A * B + beta * C Currently below cases are supported

Gets the random number generator.

Relu means rectified linear function, i.e, y = max(0, x) is applied to the CTensors elementwise. Args: x (Tensor): input tensor.

Returns repeater that broadcasts the TestEventListener events to all subscribers.

Fixed point sine and cosine. This uses the CORDIC algorithm in rotation mode. \param mz angle in [-pi/2,pi/2] as Q1.30 \param n number of iterations (

Sum of tensor elements over given axis Args: t: Singa.tensor The array_like tensor to be sumed axis: None or int or t

Gets the number of tests that should run.

function to traverse a const strided tensor object it requires an additional vector, traversal_info {0,0,0,0 ...}, comprising (x.shape().size()+2) ele

Wait for the cuda streams used by the communicator to finish their operations.

Adds a test part result to the list.

Used in the EXPECT_TRUE/FALSE(bool_expression).

Starts capturing stderr.

Clears the object.