def test_test_value_oob_gets_error_message(self):
"""The 'predict' option with the 'test_value=<INT>' flag requires that
<INT> be between 0 and 41608."""
res = predict_model.predict(test_value=-1)
assert res == self.err_msg
res = predict_model.predict(test_value=40283)
assert res == self.err_msg
def prediction():
st.title('Buy or Not Buy')
age = st.slider("Age", 18, 100)
salary = st.number_input("Salary", min_value=5000, max_value=10000000000)
label_map = {0: "Not Buying", 1: "Can Buy"}
if st.button("Classify"):
output = predict_model.predict([[age, salary]])[0]
st.write(label_map[output])
def predict_(input: Input):
probas = predict([input.text]).round(3)
response = {
'texto': input.text,
'proba_no_confiable': probas[0, 0],
'proba_confiable': probas[0, 1]
}
return response
def test_correct_test_value_call_works(self): # NOQA
"""Dumb reference test: calling predict with test_value=0 should
result in an answer of exp(8.3232). TODO: consider mocking the
calls to preprocess.preprocess() and load_learner(); might be
faster."""
res = predict_model.predict(data_path=self.data_path,
models_path=self.models_path,
test_value=0)
assert abs(float(res) - 4198.543914975132) < 0.01
def test_correct_new_value_call_works(self):
"""Dumb reference test: calling predict with new_value=<example>
should result in an answer of exp(8.3232). TODO: consider mocking
out the calls to preprocess.preprocess() and load_learner(); might be
faster."""
# Fake a test_value from the existing pre-made dataframe
# Use .iloc[0] to make sure we're using a Series per expectations
res = predict_model.predict(data_path=self.data_path,
models_path=self.models_path,
new_value='example_data_row.csv')
assert abs(float(res) - 4198.543914975132) < 0.01
def test_correct_test_value_call_with_context_works(self):
"""Dumb reference test: calling predict with test_value=0 and
context=True should result in an answer of exp(8.3232), with
appropriate context. TODO: consider mocking out the calls to
preprocess.preprocess() and load_learner(); might be faster."""
res = predict_model.predict(data_path=self.data_path,
models_path=self.models_path,
test_value=0,
context=True)
assert res == ('The predicted value is 4198.543914975132 and the '
'actual value is 4097.0.')
def predict_beer_type(request_body: beerType):
"""
Returns a single beer style prediction.
Parameters
----------
* **review_aroma** (float): Score given by reviewer regarding beer aroma- ranges from 1 to 5.
* **review_appearance** (float): Score given by reviewer regarding beer appearance - ranges from 1 to 5.
* **review_palate** (float): Score given by reviewer regarding beer palate- ranges from 1 to 5.
* **review_taste** (float): Score given by reviewer regarding beer taste - ranges from 1 to 5.
* **brwery_name** (string): Name of brewery - ranges from 1 to 5.
"""
data = create_beer_dataset(request_body, ohe, scaler)
device = get_device()
prediction = predict(data, model, device, y_encoder)
print(prediction)
return prediction
def postJsonHandler():
content = request.get_json()
user_id = "karunr" #tmp fix
user_folder = str.format("data/external/users/{0}", user_id)
if not os.path.isdir(user_folder):
os.makedirs(user_folder)
user_file_path = str.format("data/external/users/{0}/{1}_{2}.json",
user_id, user_id, int(time.time()))
csv_file_path = str.format("data/external/users/{0}/{1}_{2}.csv", user_id,
user_id, int(time.time()))
with open(user_file_path, 'w') as outfile:
json.dump(content, outfile)
parse_json(user_file_path, csv_file_path)
incoming_df = pd.read_csv(csv_file_path)
processed_df = process_incoming_data(incoming_df)
print(processed_df.columns)
ml_model = load_model()
output = predict(ml_model, processed_df)
print(output)
return output[0]
def service():
'''Serve a html document returning a sorted table with the score for each category.'''
# TODO: Add json serialized output
try:
# parse argument
query = request.args.get('q')
cat_only = request.args.get('cat_only')
# predict category
test_text, cats = predict(query,nlp=nlp)
# store categories into dataframe
cats_df=pd.Series(cats,name="Score").sort_values(ascending=False).to_frame()*100
# flow control for testing
if cat_only == "yes":
cat = cats_df['Score'].idxmax()
return cat
# create the categories html table
cats_str=cats_df.to_html(float_format='{:.0f}%'.format)
except:
# TODO: Catch exception by type to specialized the server error message.
cats_str="Server error!"
output = f'<!DOCTYPE html><body>{test_text}<P>{cats_str}:</body></html>'
return output
debug = False
# set paths
BASE_PATH = 'data/raw/training_data'
BASE_PATH_NORMALIZED = 'data/processed/normalized-images'
if args.prediction:
# use main to predict a serie of test files. See predict doc-string for more info
assert args.image != '', "no path prediction image, use -im"
# get model dir
image_list = os.listdir(args.image)
if args.model == 'default':
predict_model_path = 'models/default.h5'
else:
predict_model_path = args.model
# predict
p_model = get_model()
p_model.load_weights(predict_model_path)
predict(p_model, image_list)
# find saved files in 'models/predictions'
else:
# generate training and validation data
verse_dataset = VerseDataset(base_path=BASE_PATH_NORMALIZED, seed=args.seed, split=args.fraction, debug=debug)
training_dataset = verse_dataset.get_dataset('train').batch(args.batch_size)
validation_dataset = verse_dataset.get_dataset('validation').batch(args.batch_size)
train_u_net(training_dataset, validation_dataset, args.epochs, args.steps_per_epoch)
print("--DONE--")
def create_prediction():
data = request.data or '{}'
body = json.loads(data)
return jsonify(predict(body))
def predict():
if request.method == 'POST':
age = request.form['age']
salary = request.form['salary']
output = predict_model.predict([[age, salary]])[0]
return render_template('index.html', age=age, salary=salary, output=output)
def index():
output = predict_model.predict([[15, 150000]])[0]
return render_template('index.html')
def test_both_parameters_gets_error_message(self):
"""The 'predict' option should either be called with the 'test_value=
<INT>' flag, or with the 'new_value=<FILENAME>' flag - but not both"""
res = predict_model.predict(test_value=42, new_value='fake_file.csv')
assert res == self.err_msg
def test_no_parameters_gets_error_message(self):
"""The 'predict' option should either be called with the 'test_value=
<INT>' flag, or with the 'new_value=<FILENAME>' flag"""
res = predict_model.predict()
assert res == self.err_msg