asd

app.py

@@ -52,9 +52,8 @@
 # import dependencies
 import gradio as gr
 from openai import OpenAI
-import os
-import re
-from transformers import pipeline, DistilBertForSequenceClassification, DistilBertTokenizerFast
+import torch
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
 
 # define the openai key
 api_key = "sk-proj-UCoZZMs4MyfyHwXdHjT8T3BlbkFJjYkSZyPfIPNqXfXwoekm"
@@ -65,16 +64,18 @@ client = OpenAI(api_key = api_key)
 # finetuned model instance
 finetuned_model = "ft:gpt-3.5-turbo-0125:personal::9qGC8cwZ"
 
-# Load the AI detection model
-model_name = "tommyliphys/ai-detector-distilbert"
-model = DistilBertForSequenceClassification.from_pretrained(model_name)
-tokenizer = DistilBertTokenizerFast.from_pretrained(model_name)
-
-pipe = pipeline("text-classification", model=model, tokenizer=tokenizer)
+# Load model directly
+tokenizer = AutoTokenizer.from_pretrained("tommyliphys/ai-detector-distilbert")
+model = AutoModelForSequenceClassification.from_pretrained("tommyliphys/ai-detector-distilbert")
 
 # Define the function to get predictions
 def get_prediction(text):
-    return pipe(text)[0]
+    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)
+    with torch.no_grad():
+        outputs = model(**inputs)
+    probabilities = torch.nn.functional.softmax(outputs.logits, dim=-1)
+    ai_probability = probabilities[0][1].item()  # Assuming 1 is the index for "AI"
+    return {"label": "AI" if ai_probability > 0.5 else "Human", "score": ai_probability}
 
 # function to humanize the text
 def humanize_text(AI_text):