setup ai-model sentiment

config.py

@@ -1,5 +1,8 @@
-import os
+
 from dotenv import load_dotenv
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+import torch
+import os
 
 load_dotenv()
 
@@ -9,6 +12,7 @@ HF_TOKEN = os.getenv("HF_TOKEN")
 
 
 SENTIMENT_API = "https://huggingface.co/tabularisai/multilingual-sentiment-analysis"
+model_name = "tabularisai/multilingual-sentiment-analysis"
 # OBJECT_DETECTION_API = "https://api-inference.huggingface.co/models/facebook/detr-resnet-50"
 # TEXT_GEN_API = "https://api-inference.huggingface.co/models/gpt2"
 # SUMMARIZATION_API = "https://api-inference.huggingface.co/models/facebook/bart-large-cnn"

functions/sentiment.py

@@ -1,16 +1,59 @@
-import requests
-from config import SENTIMENT_API, headers
 
-async def sentiment_analysis(text:str)->str:
-    payload = {"inputs":text}
+from config import model_name
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+import torch
+
+
+# import os
+# # Отключаем все прокси-переменные
+# os.environ.pop('HTTP_PROXY', None)
+# os.environ.pop('HTTPS_PROXY', None)
+# os.environ.pop('http_proxy', None)
+# os.environ.pop('https_proxy', None)
+# os.environ.pop('ALL_PROXY', None)
+
+# async def sentiment_analysis(text:str)->str:
+#     payload = {"inputs":text}
+#     try:
+#         response = requests.post(SENTIMENT_API, headers=headers, json=payload)
+#         result = response.json()
+#         if isinstance(result, list) and len(result)>0:
+#             labels = result[0]
+#             best = max(labels, key=lambda x: x['score'])
+#             label_map = {"LABEL_0":"негативный", "LABEL_1": "нейтральный", "LABEL_2": "позитивный"}
+#             return label_map.get(best['label'], best['label'])
+#         return "не удалось определить"
+#     except Exception as e:
+#         return f"Error: {str(e)}"
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForSequenceClassification.from_pretrained(model_name)
+
+async def sentiment_analysis(text)->str:
+    '''принимает строку, возвращает тональность'''
     try:
-        response = requests.post(SENTIMENT_API, headers=headers, json=payload)
-        result = response.json()
-        if isinstance(result, list) and len(result)>0:
-            labels = result[0]
-            best = max(labels, key=lambda x: x['score'])
-            label_map = {"LABEL_0":"негативный", "LABEL_1": "нейтральный", "LABEL_2": "позитивный"}
-            return label_map.get(best['label'], best['label'])
-        return "не удалось определить"
+        inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)
+        with torch.no_grad():
+            outputs = model(**inputs)
+        #вероятности классов
+        probabilities = torch.nn.functional.softmax(outputs.logits, dim=-1)
+        pred_class = torch.argmax(probabilities, dim=-1).tolist()
+
+        if pred_class <=1:
+            return "Негативный"
+        elif pred_class ==2:
+            return "Нейтральный"
+        else:
+            return "Позитивный"
+
     except Exception as e:
-        return f"Error: {str(e)}"
+        return f"Error: {str(e)}"
+#
+#
+# texts = [
+#     "I absolutely love the new design of this app!", "The customer service was disappointing.", "The weather is fine, nothing special.",
+#     "Я в восторге от этого нового гаджета!", "Этот сервис оставил у меня только разочарование.", "Встреча была обычной, ничего особенного.",
+#     ]
+#
+# for text, sentiment in zip(texts, sentiment_analysis(texts)):
+#     print(f"Text: {text}\nSentiment: {sentiment}\n")