tilelang.carver.template.conv¶

Classes¶

ConvTemplate

A template for convolution (Conv).

Module Contents¶

class tilelang.carver.template.conv.ConvTemplate¶

Bases: tilelang.carver.template.base.BaseTemplate

A template for convolution (Conv).

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

N¶

The number of input samples processed simultaneously in a batch.

Type:

int

C¶

The number of input feature maps.

Type:

int

H¶

The height of the input feature maps.

Type:

int

W¶

The width of the input feature maps.

Type:

int

F¶

The number of filters (kernels) applied, determining output depth.

Type:

int

K¶

The spatial dimensions of each convolutional filter.

Type:

int

S¶

The step size by which the kernel slides across the input.

Type:

int

D¶

The spacing between kernel elements, controlling receptive field expansion.

Type:

int

P¶

The number of pixels added to input borders to control output spatial dimensions.

Type:

int

in_dtype¶

Data type of input matrices.

Type:

str

out_dtype¶

Data type of output matrix.

Type:

str

accum_dtype¶

Data type used for accumulation.

Type:

str

with_bias¶

Whether to add a bias term.

Type:

bool

N: int¶
C: int¶
H: int¶
W: int¶
F: int¶
K: int¶
S: int¶
D: int¶
P: int¶
in_dtype: str = 'float16'¶
out_dtype: str = 'float16'¶
accum_dtype: str = 'float16'¶
with_bias: bool = False¶
get_hardware_aware_configs(arch=None, topk=10)¶

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]

initialize_function()¶

Defines and initializes the convolution computation.

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

Raises:

AssertionError – If N, C, H, W, F, K, S, D, P are not positive integers.

Return type:

None

params_as_dict()¶

Returns the template parameters as a dictionary.

Returns:

Dictionary containing template parameter values.

Return type:

dict

property class_attributes¶

Returns the class attributes in dictionary form.

Returns:

Dictionary of class attributes.

Return type:

dict

__repr__()¶

Returns a string representation of the class instance.

Returns:

A formatted string representation of the class.

Return type:

str