# Python 2 and 3 compatibility
from __future__ import print_function, absolute_import, division, unicode_literals, with_statement
# In[ ]:
# Make sure python version is compatible with pyTorch
from cleanlab.util import VersionWarning
python_version = VersionWarning(
warning_str="pyTorch supports Python version 2.7, 3.5, 3.6, 3.7.",
list_of_compatible_versions=[2.7, 3.5, 3.6],
)
# In[ ]:
if python_version.is_compatible(): # pragma: no cover
import argparse
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from torch.autograd import Variable
from torch.utils.data.sampler import SubsetRandomSampler
import numpy as np
# In[ ]:
MNIST_TRAIN_SIZE = 60000
MNIST_TEST_SIZE = 10000
labels = z_list[:num_labels]
content = " ".join(z_list[num_labels:])
for l in labels:
single_label_cook_data.append(l + " " + content)
with open(data_dir + 'cooking.train.txt', 'w') as f:
f.writelines(single_label_cook_data[:-5000])
with open(data_dir + 'cooking.test.txt', 'w') as f:
f.writelines(single_label_cook_data[-5000:])
# Clean-up download files
shutil.rmtree(data_dir + 'cooking')
if python_version.is_compatible():
from fastText import train_supervised
import cleanlab
from cleanlab.models.fasttext import FastTextClassifier
from sklearn.metrics import accuracy_score
import numpy as np
# Set-up for testing.
import os
cwd = os.getcwd()
DATA_DIR = cwd + '/fasttext_data/'
# Create train and test datasets for testing.
create_cooking_dataset(DATA_DIR)
# Load train text data