File size: 3,622 Bytes
26d5b81 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 | #include <iostream>
#include <exception>
#include "../include/neuroflow/model.hpp"
#include "../include/neuroflow/memory.hpp"
#include "../include/neuroflow/networks.hpp"
using namespace neuroflow;
int main() {
try {
std::cout << "Step by step forward test..." << std::endl;
NeuroFlowModel::Config cfg;
cfg.input_dim = 64;
cfg.hidden_dim = 32;
cfg.output_dim = 5;
cfg.memory_slots = 8;
cfg.memory_dim = 16;
cfg.num_layers = 1;
cfg.num_associations = 2;
cfg.use_mla = false;
NeuroFlowModel model(cfg);
size_t batch = 2;
Tensor x({batch, cfg.input_dim});
for (size_t i = 0; i < x.numel(); ++i) x.as_fp32()[i] = 0.1f * i;
std::cout << "1. Input projection..." << std::endl;
Tensor h = model.input_proj_linear->forward(x);
std::cout << " h shape: [" << h.shape_[0] << ", " << h.shape_[1] << "]" << std::endl;
h = model.input_proj_norm->forward(h);
h = model.input_proj_gelu->forward(h);
std::cout << " after norm/gelu: [" << h.shape_[0] << ", " << h.shape_[1] << "]" << std::endl;
std::cout << "2. SN forward..." << std::endl;
auto sn_out = model.sn->forward(h);
std::cout << " saliency: [" << sn_out.saliency.shape_[0] << ", " << sn_out.saliency.shape_[1] << "]" << std::endl;
std::cout << " gates: [" << sn_out.gates.shape_[0] << ", " << sn_out.gates.shape_[1] << "]" << std::endl;
std::cout << "3. ECN forward..." << std::endl;
auto ecn_out = model.ecn->forward(h);
std::cout << " decision: [" << ecn_out.decision.shape_[0] << ", " << ecn_out.decision.shape_[1] << "]" << std::endl;
std::cout << " value: [" << ecn_out.value.shape_[0] << ", " << ecn_out.value.shape_[1] << "]" << std::endl;
std::cout << "4. Memory encode..." << std::endl;
Tensor mem_seed = model.memory->encode(h);
std::cout << " mem_seed: [" << mem_seed.shape_[0] << ", " << mem_seed.shape_[1] << "]" << std::endl;
std::cout << "5. DMN forward..." << std::endl;
auto dmn_out = model.dmn->forward(mem_seed);
std::cout << " vision shape size: " << dmn_out.vision.shape_.size() << std::endl;
for (size_t i = 0; i < dmn_out.vision.shape_.size(); ++i)
std::cout << " dim " << i << ": " << dmn_out.vision.shape_[i] << std::endl;
std::cout << "6. Memory retrieve..." << std::endl;
auto mem_out = model.memory->forward(h);
std::cout << " retrieved: [" << mem_out.retrieved.shape_[0] << ", " << mem_out.retrieved.shape_[1] << "]" << std::endl;
std::cout << "7. Reshaping dmn_out.vision..." << std::endl;
std::cout << " vision numel: " << dmn_out.vision.numel() << std::endl;
std::cout << " trying reshape to [" << batch << ", " << dmn_out.vision.shape_[1] << "]" << std::endl;
std::cout << " expected numel: " << (batch * dmn_out.vision.shape_[1]) << std::endl;
if (dmn_out.vision.numel() != batch * dmn_out.vision.shape_[1]) {
std::cout << " MISMATCH! vision actual shape may be different" << std::endl;
}
Tensor dmn_weighted = dmn_out.vision.reshape({batch, dmn_out.vision.shape_[1]});
std::cout << " reshape success" << std::endl;
std::cout << "All steps passed!" << std::endl;
return 0;
} catch (const std::exception& e) {
std::cout << "Error: " << e.what() << std::endl;
return 1;
}
}
|