def recommend_for_user():
result_df = pd.DataFrame()
already_rated = []
recDict = {}
recList = []
k = 10
for i in range(0, len(user_map)):
client_id = user_map[i]
user_idx = np.searchsorted(user_map, client_id)
user_rated = [np.searchsorted(item_map, i) for i in already_rated]
recommendations = generate_recommendations(user_idx, user_rated,
row_factor, col_factor, k)
article_recommendations = [int(item_map[i]) for i in recommendations]
# article_recommendations = article_recommendations ## convert the array of item into list
try:
recDict = {
'clientId': client_id.decode("utf-8"),
"rec": article_recommendations
}
except:
recDict = {'clientId': client_id, "rec": article_recommendations}
recList.append(recDict)
# result = pd.DataFrame( {'clientId':client_id.decode("utf-8") , "recommendations" : article_recommendations })
# result_df = result_df.append( result )
if i % 10000 == 0: print(i)
## save the result file
with open('trainer/result/nst_recommend_for_users.json', 'w') as f:
json.dump(recList, f)
## upload the content to GCP bucket
blob.upload_from_filename('trainer/result/nst_recommend_for_users.json')
def results():
if not 'name' in session or ('name' in session and session['name'] is None):
return redirect(url_for('landing'))
title = 'Market Change Prediction Results'
text = str(request.args.get("text"))
recs = generate_recommendations(text)
form = ReturnForm()
form_name = 'Return to Dashboard'
if request.method == 'POST':
return redirect(url_for('dashboard'))
return render_template('results.html', title=title, form=form, form_name=form_name, text=text, recs=recs)
def results():
title = 'Based on your responses, we recommend'
data = request.args
recs = generate_recommendations(data)
drug_1 = recs[0]
# Score, name, description, id list
drug_1 = (int(drug_1[0][0]), drug_1[1], get_desc(drug_1[1]), drug_1[0][1])
drug_2 = recs[1]
drug_2 = (int(drug_2[0][0]), drug_2[1], get_desc(drug_2[1]), drug_2[0][1])
drug_3 = recs[2]
drug_3 = (int(drug_3[0][0]), drug_3[1], get_desc(drug_3[1]), drug_3[0][1])
return render_template('results.html', title = title, drug_1=drug_1, drug_2=drug_2, drug_3=drug_3)
def results():
title = 'Based on your responses, we recommend:'
data = request.args
recs = generate_recommendations(data)
drug_1 = recs[0][1]
drug_2 = recs[1][1]
drug_3 = recs[2][1]
return render_template('results.html',
title=title,
drug_1=drug_1,
drug_2=drug_2,
drug_3=drug_3)
def main(args):
model_dir = os.path.join(args['output_dir'])
rec_path = os.path.join(args['output_dir'])
pref_path = os.path.join(args['output_dir'])
print("The args['output_dir'] = {}".format(args['output_dir']))
if model_dir.startswith('gs://'):
print("gc path = {}".format(
os.path.join(model_dir, 'model', 'recommendation_matrix.npy')))
rec_path = os.path.join(model_dir, 'model',
'recommendation_matrix.npy')
pref_path = os.path.join(model_dir, 'model', 'Preference_matrix.npy')
else:
# model_dir here is the path for the latest recommendation matrix which was previously trained
model_dir = (os.popen("cd './jobs' && ls -t | head -1").readlines()
)[0].strip('\n')
# Loading the matrix from the cmdOutput1 path
rec_path = os.path.join('jobs', model_dir, 'model',
'recommendation_matrix.npy')
pref_path = os.path.join('jobs', model_dir, 'model',
'Preference_matrix.npy')
print("gc path 1 = {}".format(rec_path))
sess = tf.Session()
with sess.as_default():
f_r = StringIO(file_io.read_file_to_string(rec_path))
print("gc f path = {}".format(f_r))
recommendation_numpy = tf.constant(np.load(f_r),
name='recommendation_numpy')
print("gc my_variable = {}".format(recommendation_numpy.eval()))
f_p = StringIO(file_io.read_file_to_string(pref_path))
Preference_numpy = tf.constant(np.load(f_p), name='Preference_numpy')
recommendation_matrix = recommendation_numpy.eval()
preference_matrix = Preference_numpy.eval()
print("preference_matrix= {}".format(preference_matrix))
with file_io.FileIO('gs://azg_bucket/data/test.json', mode='r') as f:
data = json.load(f)
users = data["user_id"]
for user_id in users:
recommendations = generate_recommendations(
100, user_id, recommendation_matrix, preference_matrix)
print("Recommendations for user {} = {}".format(
user_id, recommendations))
import numpy as np from model import generate_recommendations client_id = 10 already_rated = [293, 294, 1000000] k = 5 m_dir = '/home/parth27/bizzbytes/tensorflow-recommendation-wals/wals_ml_engine/jobs/wals_ml_local_20210116_103341/model' user_map = np.load(m_dir+"/user.npy") item_map = np.load(m_dir+"/item.npy") row_factor = np.load(m_dir+"/row.npy") col_factor = np.load(m_dir+"/col.npy") user_idx = np.searchsorted(user_map, client_id) user_rated = [np.searchsorted(item_map, i) for i in already_rated] print(len(item_map)) recommendations = generate_recommendations(942, user_rated, row_factor, col_factor, k) article_recommendations = [item_map[i] for i in recommendations] print(article_recommendations)