| using namespace neuroflow; | |
| int main() { | |
| std::cout << "GEMM without AVX2 test..." << std::endl; | |
| size_t d_model = 64; | |
| Linear W_q(d_model, d_model, false); | |
| Tensor input({1, d_model}); | |
| float* d = input.as_fp32(); | |
| for (size_t i = 0; i < input.numel(); ++i) d[i] = 0.1f * i; | |
| Tensor x_flat = input.reshape({1, d_model}); | |
| std::cout << "x_flat shape: [" << x_flat.shape_[0] << ", " << x_flat.shape_[1] << "]" << std::endl; | |
| std::cout << "Calling W_q.forward..." << std::endl; | |
| Tensor q = W_q.forward(x_flat); | |
| std::cout << "q shape: [" << q.shape_[0] << ", " << q.shape_[1] << "]" << std::endl; | |
| std::cout << "Success!" << std::endl; | |
| return 0; | |
| } | |