Update handler.py
Browse files- handler.py +13 -4
handler.py
CHANGED
|
@@ -8,18 +8,29 @@ class EndpointHandler:
|
|
| 8 |
self.vectorizer = joblib.load(os.path.join(model_dir, 'vectorizer.joblib'))
|
| 9 |
self.model = joblib.load(os.path.join(model_dir, 'logistic_classifier.joblib'))
|
| 10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
# Verify that the tokenizer configuration is correct
|
| 12 |
with open(os.path.join(model_dir, "tokenizer.json"), "r") as file:
|
| 13 |
tokenizer_config = json.load(file)
|
| 14 |
if tokenizer_config['tokenizer'] != 'split':
|
| 15 |
raise ValueError("Tokenizer configuration does not match the expected tokenizer.")
|
| 16 |
|
|
|
|
|
|
|
| 17 |
def predict_rating(self, review):
|
| 18 |
review_tfidf = self.vectorizer.transform([review])
|
| 19 |
predicted_rating = self.model.predict(review_tfidf)[0]
|
| 20 |
return int(predicted_rating)
|
| 21 |
|
| 22 |
-
# __call__ method should be part of the class
|
| 23 |
def __call__(self, inputs):
|
| 24 |
try:
|
| 25 |
# Parse the input JSON string
|
|
@@ -41,16 +52,13 @@ class EndpointHandler:
|
|
| 41 |
if not isinstance(review, str) or not review.strip():
|
| 42 |
return json.dumps({"error": "Review must be a non-empty string."})
|
| 43 |
|
| 44 |
-
# Perform prediction
|
| 45 |
predicted_rating = self.predict_rating(review)
|
| 46 |
|
| 47 |
-
# Prepare response
|
| 48 |
response = {
|
| 49 |
"review": review,
|
| 50 |
"predicted_rating": predicted_rating
|
| 51 |
}
|
| 52 |
|
| 53 |
-
# Return JSON response
|
| 54 |
return json.dumps(response)
|
| 55 |
|
| 56 |
except json.JSONDecodeError:
|
|
@@ -58,3 +66,4 @@ class EndpointHandler:
|
|
| 58 |
|
| 59 |
except Exception as e:
|
| 60 |
return json.dumps({"error": str(e)})
|
|
|
|
|
|
| 8 |
self.vectorizer = joblib.load(os.path.join(model_dir, 'vectorizer.joblib'))
|
| 9 |
self.model = joblib.load(os.path.join(model_dir, 'logistic_classifier.joblib'))
|
| 10 |
|
| 11 |
+
# Check if the vectorizer is fitted
|
| 12 |
+
if not hasattr(self.vectorizer, 'vocabulary_'):
|
| 13 |
+
raise ValueError("The vectorizer is not fitted. Ensure the vectorizer is trained and saved correctly.")
|
| 14 |
+
|
| 15 |
+
# Check if the model is fitted
|
| 16 |
+
if not hasattr(self.model, 'classes_'):
|
| 17 |
+
raise ValueError("The model is not fitted. Ensure the model is trained and saved correctly.")
|
| 18 |
+
|
| 19 |
+
print("Vectorizer and model loaded successfully.")
|
| 20 |
+
|
| 21 |
# Verify that the tokenizer configuration is correct
|
| 22 |
with open(os.path.join(model_dir, "tokenizer.json"), "r") as file:
|
| 23 |
tokenizer_config = json.load(file)
|
| 24 |
if tokenizer_config['tokenizer'] != 'split':
|
| 25 |
raise ValueError("Tokenizer configuration does not match the expected tokenizer.")
|
| 26 |
|
| 27 |
+
print("Tokenizer configuration verified.")
|
| 28 |
+
|
| 29 |
def predict_rating(self, review):
|
| 30 |
review_tfidf = self.vectorizer.transform([review])
|
| 31 |
predicted_rating = self.model.predict(review_tfidf)[0]
|
| 32 |
return int(predicted_rating)
|
| 33 |
|
|
|
|
| 34 |
def __call__(self, inputs):
|
| 35 |
try:
|
| 36 |
# Parse the input JSON string
|
|
|
|
| 52 |
if not isinstance(review, str) or not review.strip():
|
| 53 |
return json.dumps({"error": "Review must be a non-empty string."})
|
| 54 |
|
|
|
|
| 55 |
predicted_rating = self.predict_rating(review)
|
| 56 |
|
|
|
|
| 57 |
response = {
|
| 58 |
"review": review,
|
| 59 |
"predicted_rating": predicted_rating
|
| 60 |
}
|
| 61 |
|
|
|
|
| 62 |
return json.dumps(response)
|
| 63 |
|
| 64 |
except json.JSONDecodeError:
|
|
|
|
| 66 |
|
| 67 |
except Exception as e:
|
| 68 |
return json.dumps({"error": str(e)})
|
| 69 |
+
|