add app

app.py

@@ -6,97 +6,93 @@ from transformers import AutoTokenizer, BertForSequenceClassification, AutoModel
 from transformers import TextClassificationPipeline
 import gradio as gr
 
-tokenizer = AutoTokenizer.from_pretrained('momo/KcELECTRA-base_Hate_speech_Privacy_Detection')
-model = AutoModelForSequenceClassification.from_pretrained(
-    'momo/KcELECTRA-base_Hate_speech_Privacy_Detection',
-    num_labels= 15,
-    problem_type="multi_label_classification"
-)
-
-
-pipe = TextClassificationPipeline(
-model = model,
-tokenizer = tokenizer,
-return_all_scores=True,
-function_to_apply='sigmoid'
-)
-
-def predict(text):
-  return pipe(text)[0]
-
-iface = gr.Interface(
-  fn=predict, 
-  inputs='text',
-  outputs='text',
-  examples=[["Hello! My name is Omar"]]
-)
-
-iface.launch()
-
-
-
-# # global var
-# MODEL_NAME = 'momo/KcBERT-base_Hate_speech_Privacy_Detection'
-# tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+# tokenizer = AutoTokenizer.from_pretrained('momo/KcELECTRA-base_Hate_speech_Privacy_Detection')
 # model = AutoModelForSequenceClassification.from_pretrained(
-#     MODEL_NAME,
+#     'momo/KcELECTRA-base_Hate_speech_Privacy_Detection',
 #     num_labels= 15,
 #     problem_type="multi_label_classification"
 # )
 
-# MODEL_BUF = {
-#     "name": MODEL_NAME,
-#     "tokenizer": tokenizer,
-#     "model": model,
-# }
-
-# def change_model_name(name):
-#     MODEL_BUF["name"] = name
-#     MODEL_BUF["tokenizer"] = AutoTokenizer.from_pretrained(name)
-#     MODEL_BUF["model"] = AutoModelForSequenceClassification.from_pretrained(name)
+# pipe = TextClassificationPipeline(
+# model = model,
+# tokenizer = tokenizer,
+# return_all_scores=True,
+# function_to_apply='sigmoid'
+# )
 
-# def predict(model_name, text):
-#     if model_name != MODEL_BUF["name"]:
-#         change_model_name(model_name)
-    
-#     tokenizer = MODEL_BUF["tokenizer"]
-#     model = MODEL_BUF["model"]
+# def predict(text):
+#   return pipe(text)[0]
 
-#     unsmile_labels = ["여성/가족","남성","성소수자","인종/국적","연령","지역","종교","기타 혐오","악플/욕설","clean", 'name', 'number', 'address', 'bank', 'person']
-#     num_labels = len(unsmile_labels)
+# iface = gr.Interface(
+#   fn=predict, 
+#   inputs='text',
+#   outputs='text',
+#   examples=[["Hello! My name is Omar"]]
+# )
 
-#     model.config.id2label = {i: label for i, label in zip(range(num_labels), unsmile_labels)}
-#     model.config.label2id = {label: i for i, label in zip(range(num_labels), unsmile_labels)}
+# iface.launch()
 
-#     pipe = TextClassificationPipeline(
-#     model = model,
-#     tokenizer = tokenizer,
-#     return_all_scores=True,
-#     function_to_apply='sigmoid'
-#     )
 
-#     for result in pipe(text)[0]:
-#         output = result
 
-#     return output
+# global var
+MODEL_NAME = 'momo/KcBERT-base_Hate_speech_Privacy_Detection'
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+model = AutoModelForSequenceClassification.from_pretrained(
+    MODEL_NAME,
+    num_labels= 15,
+    problem_type="multi_label_classification"
+)
 
-# if __name__ == '__main__':
-#     text = '읿딴걸 홍볿글 읿랉곭 쌑젩낄고 앉앟있냩'
+MODEL_BUF = {
+    "name": MODEL_NAME,
+    "tokenizer": tokenizer,
+    "model": model,
+}
 
-#     model_name_list = [
-#         'momo/KcELECTRA-base_Hate_speech_Privacy_Detection',
-#         "momo/KcBERT-base_Hate_speech_Privacy_Detection",
-#     ]
+def change_model_name(name):
+    MODEL_BUF["name"] = name
+    MODEL_BUF["tokenizer"] = AutoTokenizer.from_pretrained(name)
+    MODEL_BUF["model"] = AutoModelForSequenceClassification.from_pretrained(name)
 
-#     #Create a gradio app with a button that calls predict()
-#     app = gr.Interface(
-#         fn=predict,
-#         inputs=[gr.inputs.Dropdown(model_name_list, label="Model Name"), 'text'], outputs=['label'], 
-#         examples = [[MODEL_BUF["name"], text], [MODEL_BUF["name"], "4=🦀 4≠🦀"]],
-#         title="한국어 혐오표현, 개인정보 판별기 (Korean Hate Speech and Privacy Detection)",
-#         description="Korean Hate Speech and Privacy Detection."
-#         )
-#     app.launch()
+def predict(model_name, text):
+    if model_name != MODEL_BUF["name"]:
+        change_model_name(model_name)
+    
+    tokenizer = MODEL_BUF["tokenizer"]
+    model = MODEL_BUF["model"]
+
+    unsmile_labels = ["여성/가족","남성","성소수자","인종/국적","연령","지역","종교","기타 혐오","악플/욕설","clean", 'name', 'number', 'address', 'bank', 'person']
+    num_labels = len(unsmile_labels)
+
+    model.config.id2label = {i: label for i, label in zip(range(num_labels), unsmile_labels)}
+    model.config.label2id = {label: i for i, label in zip(range(num_labels), unsmile_labels)}
+
+    pipe = TextClassificationPipeline(
+    model = model,
+    tokenizer = tokenizer,
+    return_all_scores=True,
+    function_to_apply='sigmoid'
+    )
+
+    return pipe(text)[0]
+
+if __name__ == '__main__':
+    text = '읿딴걸 홍볿글 읿랉곭 쌑젩낄고 앉앟있냩'
+
+    model_name_list = [
+        'momo/KcELECTRA-base_Hate_speech_Privacy_Detection',
+        "momo/KcBERT-base_Hate_speech_Privacy_Detection",
+    ]
+
+    #Create a gradio app with a button that calls predict()
+    app = gr.Interface(
+        fn=predict,
+        inputs=[gr.inputs.Dropdown(model_name_list, label="Model Name"), 'text'], outputs=['label'], 
+        examples = [[MODEL_BUF["name"], text], [MODEL_BUF["name"], "4=🦀 4≠🦀"]],
+        title="한국어 혐오표현, 개인정보 판별기 (Korean Hate Speech and Privacy Detection)",
+        description="Korean Hate Speech and Privacy Detection."
+        )
+    app.launch()
     
 
 # # global var