Update app.py

app.py

@@ -1,5 +1,5 @@
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForSequenceClassification
+from transformers import AutoTokenizer, AutoModel, AutoModelForSequenceClassification
 import torch
 import spacy
 import subprocess
@@ -11,7 +11,7 @@ from gensim import downloader as api
 nltk.download('wordnet')
 nltk.download('omw-1.4')
 
-# Ensure the spaCy model is installed
+# Ensure the SpaCy model is installed
 try:
     nlp = spacy.load("en_core_web_sm")
 except OSError:
@@ -24,18 +24,21 @@ word_vectors = api.load("glove-wiki-gigaword-50")
 # Check for GPU and set the device accordingly
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
-# Load AI Detector model and tokenizer from Hugging Face (e.g., GPT-Neo)
-tokenizer_ai = AutoTokenizer.from_pretrained("EleutherAI/gpt-neo-2.7B")
-model_ai = AutoModelForSequenceClassification.from_pretrained("EleutherAI/gpt-neo-2.7B").to(device)
+# Load GPT-3.5-turbo model and tokenizer from Hugging Face
+tokenizer_ai = AutoTokenizer.from_pretrained("Xenova/gpt-3.5-turbo")
+model_ai = AutoModel.from_pretrained("Xenova/gpt-3.5-turbo").to(device)
 
-# AI detection function using GPT-Neo-based model
+# AI detection function using GPT-3.5-turbo-based model
 def detect_ai_generated(text):
     inputs = tokenizer_ai(text, return_tensors="pt", truncation=True, max_length=512).to(device)
     with torch.no_grad():
         outputs = model_ai(**inputs)
-        probabilities = torch.softmax(outputs.logits, dim=1)
-    ai_probability = probabilities[0][1].item() * 100  # Probability of being AI-generated
-    return f"AI-Generated Content Probability: {ai_probability:.2f}%"
+        # Since this model does not directly output classification logits, you'll need to process the hidden states
+        # For simplicity, let's just use the first hidden state for now (you may need to adjust based on your use case)
+        hidden_state = outputs.last_hidden_state[:, 0, :]  # Use the first token's representation
+        # Example: calculate some kind of score based on the hidden state
+        score = torch.mean(hidden_state).item()
+    return f"AI-Generated Content Score: {score:.2f}"
 
 # Function to get synonyms using NLTK WordNet
 def get_synonyms_nltk(word, pos):