tilelang.carver.template.gemv module#

class tilelang.carver.template.gemv.GEMVTemplate(N: Optional[int] = None, K: Optional[int] = None, trans_B: bool = True, in_dtype: str = 'float16', out_dtype: str = 'float16', accum_dtype: str = 'float16', with_bias: bool = False)#

Bases: BaseTemplate

A template for Generalized Matrix-Vector Multiplication (GEMV).

This template defines the computation for a matrix-vector multiplication with configurable parameters such as transposition, data types, and bias addition.

K: int = None#
N: int = None#
accum_dtype: str = 'float16'#
property class_attributes#

Returns the class attributes in dictionary form.

Returns:

Dictionary of class attributes.

Return type:

dict

get_hardware_aware_configs(arch: Optional[TileDevice] = None, topk: int = 10) → List[Hint]#

Retrieves optimized hardware-aware configurations.

Parameters:

  • arch (TileDevice, optional) – The target hardware architecture.

  • topk (int, optional) – Number of top configurations to consider.

Returns:

A list of optimization hints for hardware acceleration.

Return type:

List[Hint]

in_dtype: str = 'float16'#
initialize_function() → None#

Defines and initializes the GEMV computation function.

This method sets up placeholders for input matrices, computes the matrix-vector multiplication using TVM’s compute API, and optionally applies bias and type casting.

out_dtype: str = 'float16'#
params_as_dict()#

Returns the template parameters as a dictionary.

Returns:

Dictionary containing template parameter values.

Return type:

dict

trans_B: bool = True#
with_bias: bool = False#