yzhangcs

# -*- coding: utf-8 -*-

import re
from datetime import datetime, timedelta

import arxiv


KEYS = ['adversarial', 'algebraic', 'algebratic', 'amr', 'auto-encoding', 'autoencoder', 'autoencoding', 'autoregressive',
        'backward', 'bayes', 'bayesian', 'bethe', 'bilexical', 'bipartite', 'bregman', 'carlo', 'chomsky', 'circuit', 'clique',

yzhangcs

# -*- coding: utf-8 -*-

from functools import reduce

import torch
import torch.autograd as autograd
from supar.utils.common import MIN
from torch.autograd import Function

yzhangcs

# -*- coding: utf-8 -*-

import argparse
from functools import partial
from multiprocessing.dummy import Pool

import nltk
from google_trans_new import google_translator
from tqdm import tqdm

yzhangcs

# -*- coding: utf-8 -*-

import argparse
from functools import partial
from multiprocessing.dummy import Pool

import nltk
from google_trans_new import google_translator
from tqdm import tqdm

yzhangcs

# -*- coding: utf-8 -*-

import argparse
from collections import Counter


def factorize(tags):
    spans = []
    for i, tag in enumerate(tags):
        if tag.startswith(('B', 'S')):

yzhangcs

# -*- coding: utf-8 -*-

import argparse


def bio2bioes(fin, fout):
    r'''Convert BIO file to BIOES file.
    Parameters:
        fin (str): the file to convert.
        fout (str): the file to save.

yzhangcs

# -*- coding: utf-8 -*-

import torch
import torch.nn as nn
import torch.nn.functional as F


class Biaffine(nn.Module):
    """
    Biaffine layer for first-order scoring.

yzhangcs

# -*- coding: utf-8 -*-

import torch
import torch.nn as nn


class GAT(nn.Module):

    def __init__(self, n_input, n_inner, n_layers, alpha=0.1, dropout=0.5):
        super(GAT, self).__init__()

yzhangcs

# -*- coding: utf-8 -*-

import torch
import torch.autograd as autograd
import torch.nn as nn


class MatrixTree(nn.Module):

    def __init__(self, *args, **kwargs):

yzhangcs

def tarjan(tree):
    tree[0] = -1
    # record the search order, i.e., the timestep
    dfn = [-1] * len(tree)
    # record the the smallest timestep in a SCC
    low = [-1] * len(tree)
    # push the visited into the stack
    stack, onstack = [], [False] * len(tree)

    def connect(i, timestep):