#ifndef PARAMETERS_H_
#define PARAMETERS_H_

#include "ap_fixed.h"
#include "ap_int.h"

#include "nnet_utils/nnet_code_gen.h"
#include "nnet_utils/nnet_helpers.h"
// hls-fpga-machine-learning insert includes
#include "nnet_utils/nnet_activation.h"
#include "nnet_utils/nnet_activation_stream.h"
#include "nnet_utils/nnet_conv2d.h"
#include "nnet_utils/nnet_conv2d_stream.h"
#include "nnet_utils/nnet_image.h"
#include "nnet_utils/nnet_image_stream.h"
#include "nnet_utils/nnet_merge.h"
#include "nnet_utils/nnet_merge_stream.h"
#include "nnet_utils/nnet_padding.h"
#include "nnet_utils/nnet_padding_stream.h"
#include "nnet_utils/nnet_pooling.h"
#include "nnet_utils/nnet_pooling_stream.h"
#include "nnet_utils/nnet_sepconv2d_stream.h"
#include "nnet_utils/nnet_stream.h"

// hls-fpga-machine-learning insert weights
#include "weights/w2.h"
#include "weights/b2.h"
#include "weights/w4.h"
#include "weights/b4.h"
#include "weights/w7.h"
#include "weights/b7.h"
#include "weights/w9.h"
#include "weights/b9.h"
#include "weights/w12.h"
#include "weights/b12.h"
#include "weights/w14.h"
#include "weights/b14.h"
#include "weights/w17.h"
#include "weights/b17.h"
#include "weights/w19.h"
#include "weights/b19.h"
#include "weights/w23.h"
#include "weights/b23.h"
#include "weights/w25.h"
#include "weights/b25.h"
#include "weights/w29.h"
#include "weights/b29.h"
#include "weights/w31.h"
#include "weights/b31.h"
#include "weights/w35.h"
#include "weights/b35.h"
#include "weights/w37.h"
#include "weights/b37.h"
#include "weights/w39.h"
#include "weights/b39.h"


// hls-fpga-machine-learning insert layer-config
// zp2d_conv1_0
struct config44 : nnet::padding2d_config {
    static const unsigned in_height = 64;
    static const unsigned in_width = 64;
    static const unsigned n_chan = 1;
    static const unsigned out_height = 66;
    static const unsigned out_width = 66;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// conv1_0
struct config2_mult : nnet::dense_config {
    static const unsigned n_in = 9;
    static const unsigned n_out = 8;
    static const unsigned reuse_factor = 72;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef conv1_0_accum_t accum_t;
    typedef conv1_0_bias_t bias_t;
    typedef conv1_0_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config2 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 66;
    static const unsigned in_width = 66;
    static const unsigned n_chan = 1;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 8;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 64;
    static const unsigned out_width = 64;
    static const unsigned reuse_factor = 72;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 66;
    static const unsigned min_width = 66;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 4096;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef conv1_0_accum_t accum_t;
    typedef conv1_0_bias_t bias_t;
    typedef conv1_0_weight_t weight_t;
    typedef config2_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config2::filt_height * config2::filt_width> config2::pixels[] = {0};

// conv1_1
struct relu_config3 : nnet::activ_config {
    static const unsigned n_in = 32768;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 160;
    typedef conv1_1_table_t table_t;
};

// zp2d_conv1_2
struct config45 : nnet::padding2d_config {
    static const unsigned in_height = 64;
    static const unsigned in_width = 64;
    static const unsigned n_chan = 8;
    static const unsigned out_height = 66;
    static const unsigned out_width = 66;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// conv1_2
struct config4_mult : nnet::dense_config {
    static const unsigned n_in = 72;
    static const unsigned n_out = 8;
    static const unsigned reuse_factor = 144;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef conv1_2_accum_t accum_t;
    typedef conv1_2_bias_t bias_t;
    typedef conv1_2_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config4 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 66;
    static const unsigned in_width = 66;
    static const unsigned n_chan = 8;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 8;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 64;
    static const unsigned out_width = 64;
    static const unsigned reuse_factor = 144;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 66;
    static const unsigned min_width = 66;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 4096;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef conv1_2_accum_t accum_t;
    typedef conv1_2_bias_t bias_t;
    typedef conv1_2_weight_t weight_t;
    typedef config4_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config4::filt_height * config4::filt_width> config4::pixels[] = {0};

// conv1_3
struct relu_config5 : nnet::activ_config {
    static const unsigned n_in = 32768;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 160;
    typedef conv1_3_table_t table_t;
};

// pool1
struct config6 : nnet::pooling2d_config {
    static const unsigned in_height = 64;
    static const unsigned in_width = 64;
    static const unsigned n_filt = 8;
    static const unsigned stride_height = 2;
    static const unsigned stride_width = 2;
    static const unsigned pool_height = 2;
    static const unsigned pool_width = 2;

    static const unsigned filt_height = pool_height;
    static const unsigned filt_width = pool_width;
    static const unsigned n_chan = n_filt;

    static const unsigned out_height = 32;
    static const unsigned out_width = 32;
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const bool count_pad = true;
    static const nnet::Pool_Op pool_op = nnet::Max;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned reuse_factor = 160;
    typedef pool1_accum_t accum_t;
};

// zp2d_conv2_0
struct config46 : nnet::padding2d_config {
    static const unsigned in_height = 32;
    static const unsigned in_width = 32;
    static const unsigned n_chan = 8;
    static const unsigned out_height = 34;
    static const unsigned out_width = 34;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// conv2_0
struct config7_mult : nnet::dense_config {
    static const unsigned n_in = 72;
    static const unsigned n_out = 16;
    static const unsigned reuse_factor = 144;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef conv2_0_accum_t accum_t;
    typedef conv2_0_bias_t bias_t;
    typedef conv2_0_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config7 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 34;
    static const unsigned in_width = 34;
    static const unsigned n_chan = 8;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 16;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 32;
    static const unsigned out_width = 32;
    static const unsigned reuse_factor = 144;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 34;
    static const unsigned min_width = 34;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 1024;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef conv2_0_accum_t accum_t;
    typedef conv2_0_bias_t bias_t;
    typedef conv2_0_weight_t weight_t;
    typedef config7_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config7::filt_height * config7::filt_width> config7::pixels[] = {0};

// conv2_1
struct relu_config8 : nnet::activ_config {
    static const unsigned n_in = 16384;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 160;
    typedef conv2_1_table_t table_t;
};

// zp2d_conv2_2
struct config47 : nnet::padding2d_config {
    static const unsigned in_height = 32;
    static const unsigned in_width = 32;
    static const unsigned n_chan = 16;
    static const unsigned out_height = 34;
    static const unsigned out_width = 34;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// conv2_2
struct config9_mult : nnet::dense_config {
    static const unsigned n_in = 144;
    static const unsigned n_out = 16;
    static const unsigned reuse_factor = 144;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef conv2_2_accum_t accum_t;
    typedef conv2_2_bias_t bias_t;
    typedef conv2_2_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_leq_nin<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config9 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 34;
    static const unsigned in_width = 34;
    static const unsigned n_chan = 16;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 16;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 32;
    static const unsigned out_width = 32;
    static const unsigned reuse_factor = 144;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 34;
    static const unsigned min_width = 34;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 1024;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef conv2_2_accum_t accum_t;
    typedef conv2_2_bias_t bias_t;
    typedef conv2_2_weight_t weight_t;
    typedef config9_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config9::filt_height * config9::filt_width> config9::pixels[] = {0};

// conv2_3
struct relu_config10 : nnet::activ_config {
    static const unsigned n_in = 16384;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 160;
    typedef conv2_3_table_t table_t;
};

// pool2
struct config11 : nnet::pooling2d_config {
    static const unsigned in_height = 32;
    static const unsigned in_width = 32;
    static const unsigned n_filt = 16;
    static const unsigned stride_height = 2;
    static const unsigned stride_width = 2;
    static const unsigned pool_height = 2;
    static const unsigned pool_width = 2;

    static const unsigned filt_height = pool_height;
    static const unsigned filt_width = pool_width;
    static const unsigned n_chan = n_filt;

    static const unsigned out_height = 16;
    static const unsigned out_width = 16;
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const bool count_pad = true;
    static const nnet::Pool_Op pool_op = nnet::Max;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned reuse_factor = 160;
    typedef pool2_accum_t accum_t;
};

// zp2d_conv3_0
struct config48 : nnet::padding2d_config {
    static const unsigned in_height = 16;
    static const unsigned in_width = 16;
    static const unsigned n_chan = 16;
    static const unsigned out_height = 18;
    static const unsigned out_width = 18;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// conv3_0
struct config12_mult : nnet::dense_config {
    static const unsigned n_in = 144;
    static const unsigned n_out = 32;
    static const unsigned reuse_factor = 576;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef conv3_0_accum_t accum_t;
    typedef conv3_0_bias_t bias_t;
    typedef conv3_0_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config12 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 18;
    static const unsigned in_width = 18;
    static const unsigned n_chan = 16;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 32;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 16;
    static const unsigned out_width = 16;
    static const unsigned reuse_factor = 576;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 18;
    static const unsigned min_width = 18;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 256;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef conv3_0_accum_t accum_t;
    typedef conv3_0_bias_t bias_t;
    typedef conv3_0_weight_t weight_t;
    typedef config12_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config12::filt_height * config12::filt_width> config12::pixels[] = {0};

// conv3_1
struct relu_config13 : nnet::activ_config {
    static const unsigned n_in = 8192;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 512;
    typedef conv3_1_table_t table_t;
};

// zp2d_conv3_2
struct config49 : nnet::padding2d_config {
    static const unsigned in_height = 16;
    static const unsigned in_width = 16;
    static const unsigned n_chan = 32;
    static const unsigned out_height = 18;
    static const unsigned out_width = 18;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// conv3_2
struct config14_mult : nnet::dense_config {
    static const unsigned n_in = 288;
    static const unsigned n_out = 32;
    static const unsigned reuse_factor = 576;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef conv3_2_accum_t accum_t;
    typedef conv3_2_bias_t bias_t;
    typedef conv3_2_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config14 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 18;
    static const unsigned in_width = 18;
    static const unsigned n_chan = 32;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 32;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 16;
    static const unsigned out_width = 16;
    static const unsigned reuse_factor = 576;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 18;
    static const unsigned min_width = 18;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 256;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef conv3_2_accum_t accum_t;
    typedef conv3_2_bias_t bias_t;
    typedef conv3_2_weight_t weight_t;
    typedef config14_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config14::filt_height * config14::filt_width> config14::pixels[] = {0};

// conv3_3
struct relu_config15 : nnet::activ_config {
    static const unsigned n_in = 8192;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 512;
    typedef conv3_3_table_t table_t;
};

// pool3
struct config16 : nnet::pooling2d_config {
    static const unsigned in_height = 16;
    static const unsigned in_width = 16;
    static const unsigned n_filt = 32;
    static const unsigned stride_height = 2;
    static const unsigned stride_width = 2;
    static const unsigned pool_height = 2;
    static const unsigned pool_width = 2;

    static const unsigned filt_height = pool_height;
    static const unsigned filt_width = pool_width;
    static const unsigned n_chan = n_filt;

    static const unsigned out_height = 8;
    static const unsigned out_width = 8;
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const bool count_pad = true;
    static const nnet::Pool_Op pool_op = nnet::Max;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned reuse_factor = 160;
    typedef pool3_accum_t accum_t;
};

// zp2d_conv4_0
struct config50 : nnet::padding2d_config {
    static const unsigned in_height = 8;
    static const unsigned in_width = 8;
    static const unsigned n_chan = 32;
    static const unsigned out_height = 10;
    static const unsigned out_width = 10;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// conv4_0
struct config17_mult : nnet::dense_config {
    static const unsigned n_in = 288;
    static const unsigned n_out = 64;
    static const unsigned reuse_factor = 1152;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef conv4_0_accum_t accum_t;
    typedef conv4_0_bias_t bias_t;
    typedef conv4_0_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config17 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 10;
    static const unsigned in_width = 10;
    static const unsigned n_chan = 32;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 64;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 8;
    static const unsigned out_width = 8;
    static const unsigned reuse_factor = 1152;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 10;
    static const unsigned min_width = 10;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 64;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef conv4_0_accum_t accum_t;
    typedef conv4_0_bias_t bias_t;
    typedef conv4_0_weight_t weight_t;
    typedef config17_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config17::filt_height * config17::filt_width> config17::pixels[] = {0};

// conv4_1
struct relu_config18 : nnet::activ_config {
    static const unsigned n_in = 4096;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 1024;
    typedef conv4_1_table_t table_t;
};

// zp2d_conv4_2
struct config51 : nnet::padding2d_config {
    static const unsigned in_height = 8;
    static const unsigned in_width = 8;
    static const unsigned n_chan = 64;
    static const unsigned out_height = 10;
    static const unsigned out_width = 10;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// conv4_2
struct config19_mult : nnet::dense_config {
    static const unsigned n_in = 576;
    static const unsigned n_out = 64;
    static const unsigned reuse_factor = 1152;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef conv4_2_accum_t accum_t;
    typedef conv4_2_bias_t bias_t;
    typedef conv4_2_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config19 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 10;
    static const unsigned in_width = 10;
    static const unsigned n_chan = 64;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 64;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 8;
    static const unsigned out_width = 8;
    static const unsigned reuse_factor = 1152;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 10;
    static const unsigned min_width = 10;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 64;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef conv4_2_accum_t accum_t;
    typedef conv4_2_bias_t bias_t;
    typedef conv4_2_weight_t weight_t;
    typedef config19_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config19::filt_height * config19::filt_width> config19::pixels[] = {0};

// conv4_3
struct relu_config20 : nnet::activ_config {
    static const unsigned n_in = 4096;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 1024;
    typedef conv4_3_table_t table_t;
};

// up3
struct config21 : nnet::resize_config {
    static const unsigned height = 8;
    static const unsigned width = 8;
    static const unsigned n_chan = 64;
    static const unsigned new_height = 16;
    static const unsigned new_width = 16;
};

// cat
struct config22 : nnet::concat_config {
    static const unsigned n_elem1_0 = 16;
    static const unsigned n_elem1_1 = 16;
    static const unsigned n_elem1_2 = 64;
    static const unsigned n_elem2_0 = 16;
    static const unsigned n_elem2_1 = 16;
    static const unsigned n_elem2_2 = 32;

    static const int axis = 3;
};

// zp2d_deconv3_0
struct config52 : nnet::padding2d_config {
    static const unsigned in_height = 16;
    static const unsigned in_width = 16;
    static const unsigned n_chan = 96;
    static const unsigned out_height = 18;
    static const unsigned out_width = 18;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// deconv3_0
struct config23_mult : nnet::dense_config {
    static const unsigned n_in = 864;
    static const unsigned n_out = 32;
    static const unsigned reuse_factor = 864;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef deconv3_0_accum_t accum_t;
    typedef deconv3_0_bias_t bias_t;
    typedef deconv3_0_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_leq_nin<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config23 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 18;
    static const unsigned in_width = 18;
    static const unsigned n_chan = 96;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 32;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 16;
    static const unsigned out_width = 16;
    static const unsigned reuse_factor = 864;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 18;
    static const unsigned min_width = 18;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 256;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef deconv3_0_accum_t accum_t;
    typedef deconv3_0_bias_t bias_t;
    typedef deconv3_0_weight_t weight_t;
    typedef config23_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config23::filt_height * config23::filt_width> config23::pixels[] = {0};

// deconv3_1
struct relu_config24 : nnet::activ_config {
    static const unsigned n_in = 8192;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 1024;
    typedef deconv3_1_table_t table_t;
};

// zp2d_deconv3_2
struct config53 : nnet::padding2d_config {
    static const unsigned in_height = 16;
    static const unsigned in_width = 16;
    static const unsigned n_chan = 32;
    static const unsigned out_height = 18;
    static const unsigned out_width = 18;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// deconv3_2
struct config25_mult : nnet::dense_config {
    static const unsigned n_in = 288;
    static const unsigned n_out = 32;
    static const unsigned reuse_factor = 1152;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef deconv3_2_accum_t accum_t;
    typedef deconv3_2_bias_t bias_t;
    typedef deconv3_2_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config25 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 18;
    static const unsigned in_width = 18;
    static const unsigned n_chan = 32;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 32;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 16;
    static const unsigned out_width = 16;
    static const unsigned reuse_factor = 1152;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 18;
    static const unsigned min_width = 18;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 256;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef deconv3_2_accum_t accum_t;
    typedef deconv3_2_bias_t bias_t;
    typedef deconv3_2_weight_t weight_t;
    typedef config25_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config25::filt_height * config25::filt_width> config25::pixels[] = {0};

// deconv3_3
struct relu_config26 : nnet::activ_config {
    static const unsigned n_in = 8192;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 1024;
    typedef deconv3_3_table_t table_t;
};

// up2
struct config27 : nnet::resize_config {
    static const unsigned height = 16;
    static const unsigned width = 16;
    static const unsigned n_chan = 32;
    static const unsigned new_height = 32;
    static const unsigned new_width = 32;
};

// cat_1
struct config28 : nnet::concat_config {
    static const unsigned n_elem1_0 = 32;
    static const unsigned n_elem1_1 = 32;
    static const unsigned n_elem1_2 = 32;
    static const unsigned n_elem2_0 = 32;
    static const unsigned n_elem2_1 = 32;
    static const unsigned n_elem2_2 = 16;

    static const int axis = 3;
};

// zp2d_deconv2_0
struct config54 : nnet::padding2d_config {
    static const unsigned in_height = 32;
    static const unsigned in_width = 32;
    static const unsigned n_chan = 48;
    static const unsigned out_height = 34;
    static const unsigned out_width = 34;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// deconv2_0
struct config29_mult : nnet::dense_config {
    static const unsigned n_in = 432;
    static const unsigned n_out = 16;
    static const unsigned reuse_factor = 432;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef deconv2_0_accum_t accum_t;
    typedef deconv2_0_bias_t bias_t;
    typedef deconv2_0_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_leq_nin<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config29 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 34;
    static const unsigned in_width = 34;
    static const unsigned n_chan = 48;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 16;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 32;
    static const unsigned out_width = 32;
    static const unsigned reuse_factor = 432;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 34;
    static const unsigned min_width = 34;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 1024;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef deconv2_0_accum_t accum_t;
    typedef deconv2_0_bias_t bias_t;
    typedef deconv2_0_weight_t weight_t;
    typedef config29_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config29::filt_height * config29::filt_width> config29::pixels[] = {0};

// deconv2_1
struct relu_config30 : nnet::activ_config {
    static const unsigned n_in = 16384;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 512;
    typedef deconv2_1_table_t table_t;
};

// zp2d_deconv2_2
struct config55 : nnet::padding2d_config {
    static const unsigned in_height = 32;
    static const unsigned in_width = 32;
    static const unsigned n_chan = 16;
    static const unsigned out_height = 34;
    static const unsigned out_width = 34;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// deconv2_2
struct config31_mult : nnet::dense_config {
    static const unsigned n_in = 144;
    static const unsigned n_out = 16;
    static const unsigned reuse_factor = 576;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef deconv2_2_accum_t accum_t;
    typedef deconv2_2_bias_t bias_t;
    typedef deconv2_2_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config31 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 34;
    static const unsigned in_width = 34;
    static const unsigned n_chan = 16;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 16;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 32;
    static const unsigned out_width = 32;
    static const unsigned reuse_factor = 576;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 34;
    static const unsigned min_width = 34;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 1024;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef deconv2_2_accum_t accum_t;
    typedef deconv2_2_bias_t bias_t;
    typedef deconv2_2_weight_t weight_t;
    typedef config31_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config31::filt_height * config31::filt_width> config31::pixels[] = {0};

// deconv2_3
struct relu_config32 : nnet::activ_config {
    static const unsigned n_in = 16384;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 512;
    typedef deconv2_3_table_t table_t;
};

// up1
struct config33 : nnet::resize_config {
    static const unsigned height = 32;
    static const unsigned width = 32;
    static const unsigned n_chan = 16;
    static const unsigned new_height = 64;
    static const unsigned new_width = 64;
};

// cat_2
struct config34 : nnet::concat_config {
    static const unsigned n_elem1_0 = 64;
    static const unsigned n_elem1_1 = 64;
    static const unsigned n_elem1_2 = 16;
    static const unsigned n_elem2_0 = 64;
    static const unsigned n_elem2_1 = 64;
    static const unsigned n_elem2_2 = 8;

    static const int axis = 3;
};

// zp2d_deconv1_0
struct config56 : nnet::padding2d_config {
    static const unsigned in_height = 64;
    static const unsigned in_width = 64;
    static const unsigned n_chan = 24;
    static const unsigned out_height = 66;
    static const unsigned out_width = 66;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// deconv1_0
struct config35_mult : nnet::dense_config {
    static const unsigned n_in = 216;
    static const unsigned n_out = 8;
    static const unsigned reuse_factor = 216;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef deconv1_0_accum_t accum_t;
    typedef deconv1_0_bias_t bias_t;
    typedef deconv1_0_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_leq_nin<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config35 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 66;
    static const unsigned in_width = 66;
    static const unsigned n_chan = 24;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 8;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 64;
    static const unsigned out_width = 64;
    static const unsigned reuse_factor = 216;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 66;
    static const unsigned min_width = 66;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 4096;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef deconv1_0_accum_t accum_t;
    typedef deconv1_0_bias_t bias_t;
    typedef deconv1_0_weight_t weight_t;
    typedef config35_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config35::filt_height * config35::filt_width> config35::pixels[] = {0};

// deconv1_1
struct relu_config36 : nnet::activ_config {
    static const unsigned n_in = 32768;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 256;
    typedef deconv1_1_table_t table_t;
};

// zp2d_deconv1_2
struct config57 : nnet::padding2d_config {
    static const unsigned in_height = 64;
    static const unsigned in_width = 64;
    static const unsigned n_chan = 8;
    static const unsigned out_height = 66;
    static const unsigned out_width = 66;
    static const unsigned pad_top = 1;
    static const unsigned pad_bottom = 1;
    static const unsigned pad_left = 1;
    static const unsigned pad_right = 1;
};

// deconv1_2
struct config37_mult : nnet::dense_config {
    static const unsigned n_in = 72;
    static const unsigned n_out = 8;
    static const unsigned reuse_factor = 288;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef deconv1_2_accum_t accum_t;
    typedef deconv1_2_bias_t bias_t;
    typedef deconv1_2_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_gt_nin_rem0<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config37 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 66;
    static const unsigned in_width = 66;
    static const unsigned n_chan = 8;
    static const unsigned filt_height = 3;
    static const unsigned filt_width = 3;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 8;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 64;
    static const unsigned out_width = 64;
    static const unsigned reuse_factor = 288;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 66;
    static const unsigned min_width = 66;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 4096;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef deconv1_2_accum_t accum_t;
    typedef deconv1_2_bias_t bias_t;
    typedef deconv1_2_weight_t weight_t;
    typedef config37_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config37::filt_height * config37::filt_width> config37::pixels[] = {0};

// deconv1_3
struct relu_config38 : nnet::activ_config {
    static const unsigned n_in = 32768;
    static const unsigned table_size = 1024;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 256;
    typedef deconv1_3_table_t table_t;
};

// final_conv
struct config58_mult : nnet::dense_config {
    static const unsigned n_in = 8;
    static const unsigned n_out = 1;
    static const unsigned reuse_factor = 8;
    static const unsigned strategy = nnet::resource;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit = DIV_ROUNDUP(n_in * n_out, reuse_factor) - n_zeros / reuse_factor;
    typedef final_conv_accum_t accum_t;
    typedef final_conv_bias_t bias_t;
    typedef final_conv_weight_t weight_t;
    template<class data_T, class res_T, class CONFIG_T>
    using kernel = nnet::DenseResource_rf_leq_nin<data_T, res_T, CONFIG_T>;
    template<class x_T, class y_T>
    using product = nnet::product::mult<x_T, y_T>;
};

struct config58 : nnet::conv2d_config {
    static const unsigned pad_top = 0;
    static const unsigned pad_bottom = 0;
    static const unsigned pad_left = 0;
    static const unsigned pad_right = 0;
    static const unsigned in_height = 64;
    static const unsigned in_width = 64;
    static const unsigned n_chan = 8;
    static const unsigned filt_height = 1;
    static const unsigned filt_width = 1;
    static const unsigned kernel_size = filt_height * filt_width;
    static const unsigned n_filt = 1;
    static const unsigned stride_height = 1;
    static const unsigned stride_width = 1;
    static const unsigned out_height = 64;
    static const unsigned out_width = 64;
    static const unsigned reuse_factor = 8;
    static const unsigned n_zeros = 0;
    static const unsigned multiplier_limit =
        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
    static const bool store_weights_in_bram = false;
    static const unsigned strategy = nnet::resource;
    static const nnet::conv_implementation implementation = nnet::conv_implementation::linebuffer;
    static const unsigned min_height = 64;
    static const unsigned min_width = 64;
    static const ap_uint<filt_height * filt_width> pixels[min_height * min_width];
    static const unsigned n_partitions = 4096;
    static const unsigned n_pixels = out_height * out_width / n_partitions;
    template<class data_T, class CONFIG_T>
    using fill_buffer = nnet::FillConv2DBuffer<data_T, CONFIG_T>;
    typedef final_conv_accum_t accum_t;
    typedef final_conv_bias_t bias_t;
    typedef final_conv_weight_t weight_t;
    typedef config58_mult mult_config;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_height = nnet::scale_index_regular<K, S, W>;
    template<unsigned K, unsigned S, unsigned W>
    using scale_index_width = nnet::scale_index_regular<K, S, W>;
};
const ap_uint<config58::filt_height * config58::filt_width> config58::pixels[] = {0};

// sigmoid
struct sigmoid_config40 : nnet::activ_config {
    static const unsigned n_in = 4096;
    static const unsigned table_size = 2048;
    static const unsigned io_type = nnet::io_stream;
    static const unsigned reuse_factor = 160;
    typedef sigmoid_table_t table_t;
};



#endif