Python-PyTorch实现基于Transformer的神经机器翻译

#include <pybind11/pybind11.h> #include <pybind11/stl.h> #include <vector> using namespace ::std; vector<vector<uint32_t>> edit_distance2_with_dp(vector<uint32_t> &x, vector<uint32_t> &y) { uint32_t lx = x.size(); uint32_t ly = y.size(); vector<vector<uint32_t>> d(lx + 1, vector<uint32_t>(ly + 1)); for (uint32_t i = 0; i < lx + 1; i++) { d[i][0] = i; } for (uint32_t j = 0; j < ly + 1; j++) { d[0][j] = j; } for (uint32_t i = 1; i < lx + 1; i++) { for (uint32_t j = 1; j < ly + 1; j++) { d[i][j] = min(min(d[i - 1][j], d[i][j - 1]) + 1, d[i - 1][j - 1] + 2 * (x[i - 1] == y[j - 1] ? 0 : 1)); } } return d; } vector<vector<uint32_t>> edit_distance2_backtracking( vector<vector<uint32_t>> &d, vector<uint32_t> &x, vector<uint32_t> &y, uint32_t terminal_symbol) { vector<uint32_t> seq; vector<vector<uint32_t>> edit_seqs(x.size() + 2, vector<uint32_t>()); /* edit_seqs: 0~x.size() cell is the insertion sequences last cell is the delete sequence */ if (x.size() == 0) { edit_seqs[0] = y; return edit_seqs; } uint32_t i = d.size() - 1; uint32_t j = d[0].size() - 1; while ((i >= 0) && (j >= 0)) { if ((i == 0) && (j == 0)) break; if ((j > 0) && (d[i][j - 1] < d[i][j])) { seq.push_back(1); // insert seq.push_back(y[j - 1]); j--; } else if ((i > 0) && (d[i - 1][j] < d[i][j])) { seq.push_back(2); // delete seq.push_back(x[i - 1]); i--; } else { seq.push_back(3); // keep seq.push_back(x[i - 1]); i--; j--; } } uint32_t prev_op, op, s, word; prev_op = 0, s = 0; for (uint32_t i = 0; i < seq.size() / 2; i++) { op = seq[seq.size() - 2 * i - 2]; word = seq[seq.size() - 2 * i - 1]; if (prev_op != 1) { s++; } if (op == 1) // insert { edit_seqs[s - 1].push_back(word); } else if (op == 2) // delete { edit_seqs[x.size() + 1].push_back(1); } else { edit_seqs[x.size() + 1].push_back(0); } prev_op = op; } for (uint32_t k = 0; k < edit_seqs.size(); k++) { if (edit_seqs[k].size() == 0) { edit_seqs[k].push_back(terminal_symbol); } } return edit_seqs; } vector<uint32_t> compute_ed2(vector<vector<uint32_t>> &xs, vector<vector<uint32_t>> &ys) { vector<uint32_t> distances(xs.size()); for (uint32_t i = 0; i < xs.size(); i++) { vector<vector<uint32_t>> d = edit_distance2_with_dp(xs[i], ys[i]); distances[i] = d[xs[i].size()][ys[i].size()]; } return distances; } vector<vector<vector<uint32_t>>> suggested_ed2_path( vector<vector<uint32_t>> &xs, vector<vector<uint32_t>> &ys, uint32_t terminal_symbol) { vector<vector<vector<uint32_t>>> seq(xs.size()); for (uint32_t i = 0; i < xs.size(); i++) { vector<vector<uint32_t>> d = edit_distance2_with_dp(xs[i], ys[i]); seq[i] = edit_distance2_backtracking(d, xs[i], ys[i], terminal_symbol); } return seq; } namespace py = pybind11; PYBIND11_PLUGIN(fast_editdistance) { py::module m("fast_editdistance", "Edit distance made by pybind11"); m.def("compute_ed2", &compute_ed2); m.def("suggested_ed2_path", &suggested_ed2_path); return m.ptr(); }