Update app.py

app.py

@@ -6,10 +6,17 @@ import re
 import datetime
 from transformers import AutoTokenizer
 import numpy as np
+from transformers import AutoModelForSequenceClassification
+from transformers import TFAutoModelForSequenceClassification
+from transformers import AutoConfig
+from scipy.special import softmax
 
 # Load tokenizer and sentiment model
 MODEL = "cardiffnlp/xlm-twitter-politics-sentiment"
 tokenizer = AutoTokenizer.from_pretrained(MODEL)
+config = AutoConfig.from_pretrained(MODEL)
+model = AutoModelForSequenceClassification.from_pretrained(MODEL)
+model.save_pretrained(MODEL)
 
 
 class ScorePredictor(nn.Module):
@@ -44,25 +51,21 @@ def preprocess_text(text):
 def predict_sentiment(text):
     if not text:
         return 0.0
-    encoded_input = tokenizer(
-        text.split(),
-        return_tensors='pt',
-        padding=True,
-        truncation=True,
-        max_length=512
-    )
-    input_ids, attention_mask = encoded_input["input_ids"], encoded_input["attention_mask"]
-    with torch.no_grad():
-        score = score_model(input_ids, attention_mask)[0].item()
-
-    scaled=score*100
-    k = 20   
-    midpoint = 0.7 
-
-    scaled_score = 1 / (1 + np.exp(-k * (score - midpoint)))
-    final_output = scaled_score * 100 
-
-    return final_output
+    text = preprocess(text)
+    encoded_input = tokenizer(text, return_tensors='pt')
+    output = model(**encoded_input)
+    scores = output[0][0].detach().numpy()
+    scores = softmax(scores)
+    ranking = np.argsort(scores)
+    ranking = ranking[::-1]
+    negative_index = None
+    for i in range(scores.shape[0]):
+        if config.id2label[ranking[i]] == 'negative':
+            negative_index = ranking[i]
+            break
+    negative_score = scores[negative_index]
+
+    return negative_score