Update app.py

app.py

@@ -2,16 +2,15 @@ from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 import torch, gradio as gr
 import re
 
-# --- Load Model (Option 1: FLAN-T5-Paraphraser) ---
-model_name = "alykassem/FLAN-T5-Paraphraser"
+# --- Load Model ---
+model_name = "Ateeqq/Text-Rewriter-Paraphraser"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-model = model.to(device)
+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]
@@ -22,39 +21,35 @@ def clean_sentence(sent):
         sent += "."
     return sent
 
-# --- Main function ---
-def paraphrase_fn(text, num_return_sequences=1, temperature=1.0, top_p=0.9):
+def paraphrase_fn(text, num_return_sequences=1, temperature=0.8, 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 = []
+    paraphrased_sentences = []
 
     for sent in sentences:
-        input_text = "paraphrase: " + sent + " </s>"
-        inputs = tokenizer([input_text], return_tensors="pt", truncation=True, padding=True).to(device)
+        prompt = "paraphraser: " + sent
+        inputs = tokenizer(prompt, 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),
+            top_p=top_p,
+            temperature=temperature,
+            no_repeat_ngram_size=2,
+            early_stopping=True
         )
-        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])
+        # Take the first unique paraphrase
+        decoded = tokenizer.batch_decode(outputs, skip_special_tokens=True)
+        clean = [clean_sentence(d) for d in decoded]
+        paraphrased_sentences.append(clean[0])
 
-    return " ".join(all_outputs).strip()
+    return " ".join(paraphrased_sentences)
 
 # --- Gradio Interface ---
 iface = gr.Interface(
@@ -62,12 +57,12 @@ iface = gr.Interface(
     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"),
+        gr.Slider(0.1, 1.5, step=0.1, value=0.8, label="Temperature"),
+        gr.Slider(0.6, 1.0, step=0.05, 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."
+    outputs=gr.Textbox(label="Paraphrased Text"),
+    title="Text Rewriter Paraphraser (T5-Base)",
+    description="High-quality model fine-tuned on 430K examples for natural, non-AI-detectable paraphrasing."
 )
 
 iface.launch()