args = parser.parse_args()
infolder = args.infolder
task = args.feature
recipe = args.recipe
print 'Loading dataset...'
data = ProfilingDataset(infolder)
print 'Loaded {} users...\n'.format(len(data.entries))
config = data.config
tasks = config.tasks
if task in tasks:
print ('Creating learning curves for %s task..' % task)
if not recipe:
recipe = config.recipes[task]
clf = from_recipe(recipe)
else:
clf = from_recipe(recipe)
print ('Loading recipe from file %s..' % recipe)
X, y = data.get_data(task)
# Cross validation with 100 iterations to get smoother mean test and train
# score curves, each time with 20% data randomly selected as a validation set.
cv = cross_validation.KFold(len(X), n_folds=5, random_state=0)
# cv = cross_validation.ShuffleSplit(len(X), n_iter=20, test_size=0.2,
# random_state=0)
title = 'Learning Curves from recipe %s' % recipe
plot_learning_curve(clf, title, X, y, ylim=(0.3, 1.01), cv=cv, n_jobs=-1)
plt.show()
else:
print('task "%s" does not exist - try one of the'
' following: %s' % (task, tasks))
parser.add_argument('-i',
'--input',
type=str,
required=True,
dest='infolder',
help='path to folder with pan dataset for a language')
parser.add_argument('-n',
'--numfolds',
type=int,
dest='num_folds',
default=4,
help='Number of folds to use in cross validation')
args = parser.parse_args()
infolder = args.infolder
num_folds = args.num_folds
print('Loading dataset...')
dataset = ProfilingDataset(infolder)
print('Loaded %s users...\n' % len(dataset.entries))
config = dataset.config
tasks = config.tasks
print('\n--------------- Thy time of Running ---------------')
for task in tasks:
tictac = from_recipe(config.recipes[task])
cross_val(dataset, task, tictac, num_folds)
# print results at end
print('\n--------------- Thy time of Judgement ---------------')
for message in log:
print(message)
if __name__ == '__main__':
parser = ArgumentParser(description='Train pan model on pan dataset')
parser.add_argument('-i', '--input', type=str,
required=True, dest='infolder',
help='path to folder with pan dataset for a language')
parser.add_argument('-o', '--output', type=str,
required=True, dest='outfolder',
help='path to folder where model should be written')
args = parser.parse_args()
infolder = args.infolder
outfolder = args.outfolder
dataset = ProfilingDataset(infolder)
print('Loaded {} users...\n'.format(len(dataset.entries)))
# get config
config = dataset.config
tasks = config.tasks
print('\n--------------- Thy time of Running ---------------')
all_models = {}
for task in tasks:
print('Learning to judge %s..' % task)
# load data
X, y = dataset.get_data(task)
tictac = from_recipe(config.recipes[task])
all_models[task] = tictac.fit(X, y)
modelfile = os.path.join(outfolder, '%s.bin' % dataset.lang)
print('Writing model to {}'.format(modelfile))
joblib.dump(all_models, modelfile, compress=3)
args = parser.parse_args()
infolder = args.infolder
task = args.feature
recipe = args.recipe
print 'Loading dataset...'
data = ProfilingDataset(infolder)
print 'Loaded {} users...\n'.format(len(data.entries))
config = data.config
tasks = config.tasks
if task in tasks:
print('Creating learning curves for %s task..' % task)
if not recipe:
recipe = config.recipes[task]
clf = from_recipe(recipe)
else:
clf = from_recipe(recipe)
print('Loading recipe from file %s..' % recipe)
X, y = data.get_data(task)
# Cross validation with 100 iterations to get smoother mean test and train
# score curves, each time with 20% data randomly selected as a validation set.
cv = cross_validation.KFold(len(X), n_folds=5, random_state=0)
# cv = cross_validation.ShuffleSplit(len(X), n_iter=20, test_size=0.2,
# random_state=0)
title = 'Learning Curves from recipe %s' % recipe
plot_learning_curve(clf,
title,
X,
y,
ylim=(0.3, 1.01),
# get config
config = dataset.config
tasks = config.tasks
print('\n--------------- Thy time of Running ---------------')
list_model_names = ['tictac', 'lda', 'voting']
total_model = {}
for model_name in list_model_names:
all_models = {}
if model_name != 'voting':
for task in tasks:
print('Learning to judge %s with %s' % (task, model_name))
# load data
X, y = dataset.get_data(task)
if 'meta' in list_model_names:
X, X_cv, y, y_cv = train_test_split(X, y, test_size=split, random_state=42, stratify=y)
tictac = from_recipe(config.recipes[task + '-' + model_name])
outline = ""
for step in tictac.steps:
if step[0] == "features":
# print type(step[1])
for tf in step[1].transformer_list:
# print type(tf[1])
# print type(tf[1].get_params())
outline += tf[0] + " with Params:[" + str(tf[1].get_params()) + "]+"
else:
# if hasattr(step[1], 'get_params'):
# outline += step[0] + " with Params:[" + str(step[1].get_params()) + "]+"
# else:
# outline += step[0]+ "+"
outline += step[0] + "+"
outline = outline[:-1] + "\n"
from argparse import ArgumentParser
from tictacs import from_recipe
if __name__ == '__main__':
parser = ArgumentParser(description='A tester of recipes for tic tacs')
parser.add_argument('--recipe', '-r', required=True, dest='recipe',
help='Path to the file where the recipe to create '
'delicious tictacs resides. All recipes must be '
'written in yaml format.',
default='recipes/example.py')
args = parser.parse_args()
recipe = args.recipe
print('Using recipe from file: %s' % recipe)
texts = ['@sly_pedantic_octopus @glorified_ml I walked on ice yesterday,'
' but no one laughed when it broke #lol',
'@blue_world I hate java #java #programming #hell',
'@BarnieTheDinosaur raaawwwwr.',
'omg that just happened. #omg #rofl #yolo'
]
print('Creating model...')
tictac = from_recipe(recipe)
print('Fitting model...')
tictac.fit(texts, [0, 1, 1, 1])
print('Predicting with model...')
res = tictac.predict(['#dog wtf omg java', '@blue yes broke #lol'])
print('Predicted %s' % res)
dest='num_folds', default=4,
help='Number of folds to use in cross validation')
args = parser.parse_args()
X_path = args.x_path
y_path = args.y_path
num_folds = args.num_folds
# This part for tira-io
with open(X_path, 'r') as xin:
X = pickle.load(xin)
with open(y_path, 'r') as yin:
y = pickle.load(yin)
######
print('Number of docs: %d,%d' % (len(X), len(y)))
for task in ['gender']:
tictac = from_recipe("./config/recipes/gender.yml")
outline = ""
for step in tictac.steps:
if step[0] == "features":
# print type(step[1])
for tf in step[1].transformer_list:
# print type(tf[1])
# print type(tf[1].get_params())
outline += tf[0] + " with Params:[" + str(tf[1].get_params()) + "]+"
else:
# if hasattr(step[1], 'get_params'):
# outline += step[0] + " with Params:[" + str(step[1].get_params()) + "]+"
# else:
# outline += step[0]+ "+"
outline += step[0] + "+"
outline = outline[:-1] + "\n"
