Update app.py

app.py

@@ -1,58 +1,73 @@
-import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
-import torch
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+import torch, gradio as gr
+import re
 
-# ✅ Free, open model (no gated access required)
-model_name = "HuggingFaceH4/zephyr-7b-alpha"
+# --- Load Model (Option 1: FLAN-T5-Paraphraser) ---
+model_name = "alykassem/FLAN-T5-Paraphraser"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
 
-model = AutoModelForCausalLM.from_pretrained(
-    model_name,
-    device_map="auto",          # auto = CPU in free Spaces
-    low_cpu_mem_usage=True
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model = model.to(device)
+model.eval()
+
+# --- Helpers ---
+def split_sentences(text):
+    sentences = re.split(r'(?<=[.!?])\s+', text.strip())
+    return [s for s in sentences if s]
+
+def clean_sentence(sent):
+    sent = re.sub(r'\s+', ' ', sent).strip()
+    if not sent.endswith(('.', '!', '?')):
+        sent += "."
+    return sent
+
+# --- Main function ---
+def paraphrase_fn(text, num_return_sequences=1, temperature=1.0, top_p=0.9):
+    if not text.strip():
+        return "⚠️ Please enter some text"
+
+    num_return_sequences = int(num_return_sequences)
+    sentences = split_sentences(text)
+    all_outputs = []
+
+    for sent in sentences:
+        input_text = "paraphrase: " + sent + " </s>"
+        inputs = tokenizer([input_text], return_tensors="pt", truncation=True, padding=True).to(device)
+
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=128,
+            num_return_sequences=num_return_sequences,
+            do_sample=True,
+            top_p=float(top_p),
+            temperature=float(temperature),
+        )
+        decoded = tokenizer.batch_decode(outputs, skip_special_tokens=True)
+
+        seen, unique = set(), []
+        for d in decoded:
+            d = clean_sentence(d)
+            if d not in seen:
+                unique.append(d)
+                seen.add(d)
+
+        all_outputs.append(unique[0])
+
+    return " ".join(all_outputs).strip()
+
+# --- Gradio Interface ---
+iface = gr.Interface(
+    fn=paraphrase_fn,
+    inputs=[
+        gr.Textbox(lines=8, placeholder="Paste text here..."),
+        gr.Slider(1, 3, step=1, value=1, label="Variants"),
+        gr.Slider(0.5, 2.0, step=0.1, value=1.0, label="Temperature"),
+        gr.Slider(0.6, 1.0, step=0.01, value=0.9, label="Top-p"),
+    ],
+    outputs=gr.Textbox(label="Output"),
+    title="📝 Writenix Paraphraser (FLAN-T5)",
+    description="Paraphrasing powered by FLAN-T5, fine-tuned on high-quality datasets."
 )
 
-# Humanizer function
-def humanize_text(ai_text):
-    if not ai_text.strip():
-        return "⚠️ Please enter some text."
-    
-    # Prompt for rewriting
-    prompt = (
-        "Rewrite the following AI-generated text so it sounds natural and human-like. "
-        "Avoid robotic tone, vary sentence length, and make it engaging:\n\n"
-        f"Text: {ai_text}\n\n"
-        "Humanized version:"
-    )
-
-    inputs = tokenizer(prompt, return_tensors="pt")
-    outputs = model.generate(
-        **inputs,
-        max_new_tokens=400,
-        temperature=1.0,
-        top_p=0.95
-    )
-    humanized = tokenizer.decode(outputs[0], skip_special_tokens=True)
-
-    return humanized
-
-# Gradio UI
-with gr.Blocks() as demo:
-    gr.Markdown("## ✨ AI Text Humanizer")
-    gr.Markdown("Paste AI-generated text below, and get a natural, human-like rewrite.")
-
-    with gr.Row():
-        ai_input = gr.Textbox(label="AI-Generated Text", placeholder="Paste your AI text here...", lines=10)
-    
-    with gr.Row():
-        humanized_output = gr.Textbox(label="✅ Humanized Text", lines=10)
-
-    run_button = gr.Button("🔄 Humanize Now")
-
-    run_button.click(
-        fn=humanize_text,
-        inputs=ai_input,
-        outputs=humanized_output
-    )
-
-demo.launch()
+iface.launch()