Utility function for computing the shape resulting from a dilation.

Applies the transfer guard level within transfer_guard_lib.

r"""Random variable with Beta distribution :math:`\operatorname{Beta}(\alpha, \beta)`. The corresponding probability density function is

Convert bfloat16 to single-precision. \param value binary_t() representation of bfloat16 value \return single-precision value

broadcast src to broadcasted shape of dst_shape_var

r"""Element-wise clipping function. Clamps(limits) all elements :math:`x_i` of the input tensor :math:`x` into the range ``[ lower, upper ]``

Collect a single batch through an array and store the array data to an empty batch, the empty batch is the front batch id in free lis

get the comp node to run the checker

! * \brief latest computing sequence from this graph * * Since new computing sequence would invalidate memory layout of older * on

r"""Element-wise division. Calculates the division for each element :math:`x_i` of the input tensor :math:`x` with the respective element :ma

Convert MegBrain DType to NumPy DType descriptor, the caller receives a new reference to the descriptor.

set cpu forward in inplace mode with which cpu forward only create one thread Note: this must be set before the

set global layout transform optimization for network, global layout optimization can auto determine the layout of every operator in

Error function. \param arg function argument \return error function value of \a arg template<typename T> typename enable<expr,T>::type erf(T arg) { re

r"""Returns :class:`.tensor.core.megbrain_graph.VarNode` of type ``var_type`` that input ``var`` depands on. If ``var_type`` is None, returns all

r"""Filter Exprs by the type of ``CallFunction``. Args: func: a built-in function, such as ``F.relu``. recursive: whe

get_matmul_kern_param

get the model io information before model loaded by model path. Args: model_path: the model path to get the model IO information

get_thread_bundle

see notes on the constructor