Creational

covid_nlp/eval.py

@@ -4,6 +4,23 @@
 from sklearn.metrics import roc_auc_score, f1_score
 from farm.utils import MLFlowLogger
 
+class SingletonEval(type):
+
+    #create a metaclass 
+    #overall idea retrieved from StackOverflow https://stackoverflow.com/questions/6760685/creating-a-singleton-in-python
+    _instances = {}
+
+    def getInstance (*args,**kwargs, theClass):
+
+        if theClass not in theClass.instances:
+
+            instance = super().getInstance(*args,**kwargs)
+            theClass._instances[theClass] = instance
+
+            return theClass._instances[theClass]
+
+
+class Singleton(metaclass = SingletonEval):
 
 def eval_question_similarity(y_true, y_pred, lang, model_name, params, user=None, log_to_mlflow=True, run_name="default"):
     # basic metrics
@@ -44,7 +61,7 @@ def eval_question_similarity(y_true, y_pred, lang, model_name, params, user=None
     y_pred = [0.5] * len(y_true)
 
     # eval & track results
-    eval_question_similarity(y_true=y_true, y_pred=y_pred, lang=lang, model_name=model_name,
-                             params=params, user="malte", log_to_mlflow=log_to_mlflow, run_name=experiment_name)
+    ourSingleton = SingletonEval()
+    ourSingleton.eval_question_similarity(y_true=y_true, y_pred=y_pred, lang=lang, model_name=model_name, params=params, user="malte", log_to_mlflow=log_to_mlflow, run_name=experiment_name)
 
 

covid_nlp/language/detect_language.py

@@ -3,10 +3,34 @@
 import pycld2 as cld2
 import hmac
 from hashlib import sha1
+from datetime import datetime
 from time import time
 import requests
 
+class Algo():
+    def setAlgo(self):
+        algorithm = 'HMAC+SHA1'
+        return algorithm
+
+    def setTime(self):
+        time = datetime.now()
+        return time
+
+    def setConcat(curr_time):
+        concat = curr_time+os.environ.get('SIL_API_KEY')
+        return concat
+
+    def setConcatB(concat):
+        concatB = (concat).encode('utf-8')
+        return concatB
+
+    def setSecretB(self):
+            secretB = os.environ.get('SIL_API_SECRET').encode('utf-8')
+            return secretB
+
+
 class LanguageDetector():
+
     def __init__(self, model = 'sil'):
         self.model = model
 
@@ -20,11 +44,13 @@ def detect_lang_cld3(self, text):
         return pred.language, 100*pred.probability
 
     def detect_lang_sil(self, text):
-        algorithm = 'HMAC+SHA1'
-        curr_time = str(int(time()))
-        concat = curr_time+os.environ.get('SIL_API_KEY')
-        concatB = (concat).encode('utf-8')
-        secretB = os.environ.get('SIL_API_SECRET').encode('utf-8')
+        yes = Algo()
+    #setting these to an algorithm class
+        algorithm = yes.setAlgo()
+        curr_time = yes.setTime()
+        concat = yes.setConcat(curr_time)
+        concatB = yes.setConcatB()
+        secretB = yes.setSecretB()
         h1 = hmac.new(secretB, concatB, sha1)
         api_sig = h1.hexdigest()
         params = {'api_key': os.environ.get('SIL_API_KEY'), 'api_sig': api_sig}

covid_nlp/modeling/tfidf/tfidf_client.py

@@ -2,6 +2,7 @@
 import re
 import pickle
 import os
+import eval
 import json
 
 from sklearn.feature_extraction.text import TfidfVectorizer
@@ -13,7 +14,7 @@
 from tfidf_train import TfidfTrainer
 
 sys.path.insert(0, "./../../")
-from eval import eval_question_similarity
+from eval import ourSingleton.eval_question_similarity
 
 
 class TfidfEvaluator():
@@ -51,7 +52,7 @@ def main():
     exp_name = "tfidf_cos_sim_2"
     params = {"sp_voc": 16000, "max_ngram": 2, "remove_stopwords": 1, 
                 "data_train": "eval, scraped", "data_sp": "eval, scraped, CORD-19.200k"}
-    eval_question_similarity(y_true=y_true, y_pred=y_pred, lang="en", model_name=model_name,
+    ourSingleton.eval_question_similarity(y_true=y_true, y_pred=y_pred, lang="en", model_name=model_name,
                              params=params, user="carmen", log_to_mlflow=True, run_name=exp_name)
 
 

covid_nlp/modeling/transformer/eval_pretrained_haystack.py

@@ -1,5 +1,6 @@
 import pandas as pd
 import numpy as np
+import eval
 
 from sklearn.metrics import roc_auc_score
 from farm.utils import MLFlowLogger
@@ -38,7 +39,7 @@ def eval_pretrained_transformers(eval_file, lang, models, pooling_methods, extra
                 df["pred"] = np.diag(cosine_similarity(res1, res2))
 
                 # eval & track results
-                eval_question_similarity(y_true=y_true, y_pred=df["pred"].values, lang=lang, model_name=model_name,
+                ourSingleton.eval_question_similarity(y_true=y_true, y_pred=df["pred"].values, lang=lang, model_name=model_name,
                                          params=params, user="malte", log_to_mlflow=log_to_mlflow, run_name=experiment_name)
 
 if __name__ == "__main__":