korakot

# authenticate
from google.colab import auth
auth.authenticate_user()
import gspread
from oauth2client.client import GoogleCredentials as GC
gc = gspread.authorize(GC.get_application_default())
# create, and save df
from gspread_dataframe import set_with_dataframe
title = 'New Sheet'
gc.create(title)  # if not exist

yzh119

from __future__ import print_function
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable

def sample_gumbel(shape, eps=1e-20):
    U = torch.rand(shape).cuda()
    return -Variable(torch.log(-torch.log(U + eps) + eps))

neubig

"""
DyNet implementation of a sequence labeler (POS taggger).
This is a translation of this tagger in PyTorch: https://gist.github.com/hal3/8c170c4400576eb8d0a8bd94ab231232

Basic architecture:
 - take words
 - run though bidirectional GRU
 - predict labels one word at a time (left to right), using a recurrent neural network "decoder"
The decoder updates hidden state based on:
 - most recent word

noisychannel

#!/usr/bin/env python

# NOTE : This script assumes that the aligments are in the src-tgt format

import optparse
import pprint
import sys
import numpy as np

optparser = optparse.OptionParser()

mjpost

#!/usr/bin/env python
# -*- coding: utf-8 -*-

# Python *sucks* at UTF-8 (don't tell me "It's fixed in Python 3"; I don't care, plus no one uses Python 3)
# If you put this at the top of every Python script, however, it get rids of most of the headaches dealing with STDIN
# and STDOUT (basically, akin to "perl -C31"). I don't know if it's all necessary; I just know that if I put it at
# the top of my scripts, most of the problems go away, and I can stop thinking about it.

import sys
import codecs