Update app.py

app.py

@@ -1,65 +1,51 @@
 import gradio as gr
-import re
-import torch
-import asyncio
-from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
+from transformers import T5Tokenizer, T5ForConditionalGeneration, pipeline
 
-# Load T5 paraphrase model
-model_name = "Vamsi/T5_Paraphrase_Paws"
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForSeq2SeqLM.from_pretrained(model_name, torch_dtype=torch.float16)
+# Load T5-small model and tokenizer
+model_name = "t5-small"
+tokenizer = T5Tokenizer.from_pretrained(model_name)
+model = T5ForConditionalGeneration.from_pretrained(model_name)
 
-# Move model to CPU
-model.to("cpu")
-
-# Initialize paraphrase pipeline
-paraphrase_pipeline = pipeline(
-    "text2text-generation",
-    model=model,
-    tokenizer=tokenizer,
-    truncation=True
-)
-
-def split_sentences(text):
-    """Split text into sentences using regex (faster than nltk)."""
-    return re.split(r'(?<=[.!?])\s+', text.strip())
-
-async def paraphrase_text(text):
-    """Paraphrases input text asynchronously while maintaining sentence structure."""
+def generate_paraphrase(text, max_length, temperature):
+    """Generate a paraphrased version of the input text using T5-small."""
     if not text.strip():
         return "⚠️ Please enter some text to paraphrase."
-
-    sentences = split_sentences(text)
     
-    # Ensure batch processing
-    formatted_input = [f"paraphrase: {sentence} </s>" for sentence in sentences if sentence]
-    
-    # Run in async mode to handle multiple requests in parallel
-    paraphrased_results = await asyncio.to_thread(
-        paraphrase_pipeline, 
-        formatted_input,
-        max_length=80,
-        do_sample=True,  
-        temperature=0.7,  
-        top_p=0.85,  
+    input_text = f"paraphrase: {text} </s>"
+    input_ids = tokenizer.encode(input_text, return_tensors="pt")
+
+    output = model.generate(
+        input_ids,
+        max_length=max_length,
         top_k=50,
-        repetition_penalty=1.2,  
-        num_return_sequences=1,  
-        batch_size=8  # Batch processing enabled ✅
+        top_p=0.95,
+        num_return_sequences=1,
+        do_sample=True
     )
 
-    # Extract and join paraphrased sentences
-    paraphrased_sentences = [result["generated_text"] for result in paraphrased_results]
-    return " ".join(paraphrased_sentences)
-
-# Define Gradio Interface with Non-Blocking Requests
-with gr.Blocks() as demo:
-    gr.Markdown("# 🚀 Fast & Parallel T5 Paraphraser")
-    input_box = gr.Textbox(label="Enter text", placeholder="Type your text to paraphrase...", lines=10)
-    output_box = gr.Textbox(label="Paraphrased Text", lines=10)
-    button = gr.Button("Paraphrase")
-
-    button.click(paraphrase_text, inputs=input_box, outputs=output_box)
+    paraphrased_text = tokenizer.decode(output[0], skip_special_tokens=True)
+    return paraphrased_text
+
+# Define Gradio Interface
+description = """
+## ✨ AI Paraphrasing Tool
+Enter a sentence and let AI generate a paraphrased version!
+- Adjust **max length** for longer outputs.
+- Tune **temperature** for more creative results.
+"""
+
+demo = gr.Interface(
+    fn=generate_paraphrase,
+    inputs=[
+        gr.Textbox(label="Enter text", placeholder="Type a sentence to paraphrase..."),
+        gr.Slider(20, 100, value=50, step=5, label="Max Output Length"),
+        gr.Slider(0.5, 1.5, value=1.0, step=0.1, label="Creativity (Temperature)"),
+    ],
+    outputs=gr.Textbox(label="Paraphrased Text"),
+    title="📝 AI Paraphraser",
+    description=description,
+    theme="huggingface",
+    live=True,
+)
 
-# ✅ Fix queue issue & enable concurrent users
-demo.queue(concurrency_count=10).launch(share=True)  # Allows 10 users to paraphrase simultaneously
+demo.launch()