Noodle
Loading...
Searching...
No Matches
ConvFile Struct Reference

File-backed convolution parameter bundle alias. More...

#include <noodle.h>

Public Attributes

uint16_t K = 3
 Kernel width, or tap count for 1D convolution.
uint16_t P = 0
 Padding per side; 65535 requests SAME-style 2D padding.
uint16_t S = 1
 Convolution stride.
uint16_t OP = 0
 User-computed output padding for transpose convolution.
const char * weight_fn = nullptr
 Weight filename.
const char * bias_fn = nullptr
 Bias filename.
Activation act = ACT_RELU
 Activation applied after adding bias.

Detailed Description

File-backed convolution parameter bundle alias.

This carries the same layout and semantics as Conv and is kept for code that prefers an explicit file-backed name.

For transpose convolution with explicit padding, callers choose OP to match the desired output width: V = (W - 1) * S - 2 * P + K + OP.

Member Data Documentation

◆ act

Activation ConvFile::act = ACT_RELU

Activation applied after adding bias.

◆ bias_fn

const char* ConvFile::bias_fn = nullptr

Bias filename.

◆ K

uint16_t ConvFile::K = 3

Kernel width, or tap count for 1D convolution.

◆ OP

uint16_t ConvFile::OP = 0

User-computed output padding for transpose convolution.

◆ P

uint16_t ConvFile::P = 0

Padding per side; 65535 requests SAME-style 2D padding.

◆ S

uint16_t ConvFile::S = 1

Convolution stride.

◆ weight_fn

const char* ConvFile::weight_fn = nullptr

Weight filename.


The documentation for this struct was generated from the following file: