Update app.py

app.py

@@ -1,13 +1,15 @@
 import gradio as gr
 import torch
 import numpy as np
-from transformers import AutoModelForSequenceClassification, AutoTokenizer
+from transformers import pipeline
 from transformers import RobertaForSequenceClassification, RobertaTokenizer
 
-# Load fine-tuned sentiment model
-sentiment_model_path = "./sentiment_model"
-sentiment_model = AutoModelForSequenceClassification.from_pretrained(sentiment_model_path)
-sentiment_tokenizer = AutoTokenizer.from_pretrained(sentiment_model_path)
+# Load fine-tuned sentiment model from Hugging Face
+sentiment_analyzer = pipeline(
+    "sentiment-analysis",
+    model="SamanthaStorm/Tether",
+    tokenizer="SamanthaStorm/Tether"
+)
 
 # Load abuse pattern model
 model_name = "SamanthaStorm/abuse-pattern-detector-v2"
@@ -50,16 +52,6 @@ EXPLANATIONS = {
     "obscure_formal": "Obscure/formal language manipulates through confusion or superiority."
 }
 
-def custom_sentiment(text):
-    inputs = sentiment_tokenizer(text, return_tensors="pt", truncation=True, padding=True)
-    with torch.no_grad():
-        outputs = sentiment_model(**inputs)
-    probs = torch.nn.functional.softmax(outputs.logits, dim=1)
-    label_idx = torch.argmax(probs).item()
-    label = sentiment_model.config.id2label[label_idx]
-    score = probs[0][label_idx].item()
-    return {"label": label, "score": score}
-
 def calculate_abuse_level(scores, thresholds):
     triggered_scores = [score for label, score in zip(LABELS, scores) if score > thresholds[label]]
     return round(np.mean(triggered_scores) * 100, 2) if triggered_scores else 0.0
@@ -76,7 +68,7 @@ def analyze_messages(input_text, risk_flags):
     if not input_text:
         return "Please enter a message for analysis."
 
-    sentiment = custom_sentiment(input_text)
+    sentiment = sentiment_analyzer(input_text)[0]
     sentiment_label = sentiment['label']
     sentiment_score = sentiment['score']
 
@@ -113,7 +105,7 @@ def analyze_messages(input_text, risk_flags):
     top_patterns = sorted(scored_patterns, key=lambda x: x[1], reverse=True)[:2]
 
     top_pattern_explanations = "\n".join([
-        f"\u2022 {label.replace('_', ' ').title()}: {EXPLANATIONS.get(label, 'No explanation available.')}"
+        f"• {label.replace('_', ' ').title()}: {EXPLANATIONS.get(label, 'No explanation available.')}"
         for label, _ in top_patterns
     ])