applications/include/network.h

Classes

	Name
struct	network_t_ This structure contains all parameters necessary for building a simple neural netowork.
struct	network_fp64_t_
struct	network_fp32_t_
struct	network_fp16_t_
struct	network_fp8_t_
struct	cnn_t_
struct	network_benchmark_t_
struct	network_single_cluster_t_

Types

	Name
typedef struct network_t_	network_t
typedef struct network_fp64_t_	network_fp64_t
typedef struct network_fp32_t_	network_fp32_t
typedef struct network_fp16_t_	network_fp16_t
typedef struct network_fp8_t_	network_fp8_t
typedef struct cnn_t_	cnn_t
typedef struct network_benchmark_t_	network_benchmark_t
typedef struct network_single_cluster_t_	network_single_cluster_t

Types Documentation

typedef network_t

typedef struct network_t_ network_t;

typedef network_fp64_t

typedef struct network_fp64_t_ network_fp64_t;

typedef network_fp32_t

typedef struct network_fp32_t_ network_fp32_t;

typedef network_fp16_t

typedef struct network_fp16_t_ network_fp16_t;

typedef network_fp8_t

typedef struct network_fp8_t_ network_fp8_t;

typedef cnn_t

typedef struct cnn_t_ cnn_t;

typedef network_benchmark_t

typedef struct network_benchmark_t_ network_benchmark_t;

typedef network_single_cluster_t

typedef struct network_single_cluster_t_ network_single_cluster_t;

Source code

// Copyright 2020 ETH Zurich and University of Bologna.
// Licensed under the Apache License, Version 2.0, see LICENSE for details.
// SPDX-License-Identifier: Apache-2.0

#pragma once

#include <stdint.h>

#include "layer.h"

typedef struct network_t_ {
    uint32_t IN_CH1;
    uint32_t IN_CH2;
    uint32_t OUT_CH;

    float *b;
    float *W;
    float *b_grad;
    float *W_grad;
    float *images;
    uint32_t *targets;

    precision_t dtype;

} network_t;

typedef struct network_fp64_t_ {
    uint32_t IN_CH1;
    uint32_t IN_CH2;
    uint32_t OUT_CH;

    double *b;
    double *W;
    double *b_grad;
    double *W_grad;

    double *images;
    uint32_t *targets;
    float learning_rate;

    precision_t dtype;

} network_fp64_t;

typedef struct network_fp32_t_ {
    uint32_t IN_CH1;
    uint32_t IN_CH2;
    uint32_t OUT_CH;

    float *b;
    float *W;
    float *b_grad;
    float *W_grad;

    float *images;
    uint32_t *targets;
    float learning_rate;

    precision_t dtype;

} network_fp32_t;

typedef struct network_fp16_t_ {
    uint32_t IN_CH1;
    uint32_t IN_CH2;
    uint32_t OUT_CH;

    __fp16 *b;
    __fp16 *W;
    __fp16 *b_grad;
    __fp16 *W_grad;

    __fp16 *images;
    uint32_t *targets;
    float learning_rate;

    precision_t dtype;

} network_fp16_t;

typedef struct network_fp8_t_ {
    uint32_t IN_CH1;
    uint32_t IN_CH2;
    uint32_t OUT_CH;

    char *b;
    char *W;
    char *b_grad;
    char *W_grad;

    char *images;
    uint32_t *targets;
    float learning_rate;

    precision_t dtype;

} network_fp8_t;

// TODO: add description for MNIST CNN network struct
typedef struct cnn_t_ {
    uint16_t CO;
    uint16_t CI;
    uint16_t H;
    uint16_t W;
    uint16_t K;
    uint16_t M;

    uint16_t padding;
    uint16_t stride;

    double *image;
    double *conv1_weights;
    double *conv1_biases;

    precision_t dtype;

} cnn_t;

typedef struct network_benchmark_t_ {
    uint32_t IN_CH;
    uint32_t OUT_CH;

    void *b;
    void *W;

    void *images;
    uint32_t *targets;

    precision_t dtype;

} network_benchmark_t;

typedef struct network_single_cluster_t_ {
    uint32_t IN_CH;
    uint32_t OUT_CH;

    void *b;
    void *W;

    void *b_grads;
    void *W_grads;

    void *images;
    uint32_t *targets;

    precision_t dtype;

} network_single_cluster_t;

---

Updated on 2023-06-19 at 09:43:56 +0000