Update app.py

app.py

@@ -1,76 +1,36 @@
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 import torch, gradio as gr
-import re
 
-# --- Load Model ---
-model_name = "Ateeqq/Text-Rewriter-Paraphraser"
+model_name = "cPower/dipper-paraphraser-xxl-tokeninc"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-model.to(device)
+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 (Batch Processing) ---
-def paraphrase_fn(text, num_return_sequences=1, temperature=0.8, top_p=0.9):
-    if not text.strip():
-        return "⚠️ Please enter some text"
-
-    sentences = split_sentences(text)
-    prompts = ["paraphraser: " + s for s in sentences]
-
-    inputs = tokenizer(
-        prompts,
-        return_tensors="pt",
-        truncation=True,
-        padding=True
-    ).to(device)
-
+def paraphrase_dipper(text, diversity=0.5, reordering=0.5):
+    inputs = tokenizer([text], return_tensors="pt", truncation=True, padding=True).to(device)
     outputs = model.generate(
         **inputs,
-        max_new_tokens=64,                # smaller for speed
-        num_return_sequences=int(num_return_sequences),
+        max_new_tokens=256,
+        top_p=1.0,
+        diversity_penalty=float(diversity),
+        num_reorder=float(reordering),
         do_sample=True,
-        top_p=top_p,
-        temperature=temperature,
-        no_repeat_ngram_size=2,
-        early_stopping=True
     )
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
-    decoded = tokenizer.batch_decode(outputs, skip_special_tokens=True)
-
-    # Since we may get multiple return sequences, group by sentence
-    paraphrased_sentences = []
-    step = int(num_return_sequences)
-    for i in range(0, len(decoded), step):
-        first_variant = clean_sentence(decoded[i])  # take the first variant only
-        paraphrased_sentences.append(first_variant)
-
-    return " ".join(paraphrased_sentences)
-
-# --- Gradio Interface ---
 iface = gr.Interface(
-    fn=paraphrase_fn,
+    fn=paraphrase_dipper,
     inputs=[
-        gr.Textbox(lines=8, placeholder="Paste text here..."),
-        gr.Slider(1, 3, step=1, value=1, label="Variants"),
-        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"),
+        gr.Textbox(lines=8, placeholder="Paste full text here..."),
+        gr.Slider(0.0, 1.0, step=0.1, value=0.5, label="Lexical Diversity"),
+        gr.Slider(0.0, 1.0, step=0.1, value=0.5, label="Reordering Amount")
     ],
-    outputs=gr.Textbox(label="Paraphrased Text"),
-    title="Text Rewriter Paraphraser (Batch Optimized)",
-    description="Fast paraphrasing powered by T5-base. Now optimized with batch processing 🚀"
+    outputs=gr.Textbox(label="Paraphrased & Humanized Text"),
+    title="DIPPER Paraphraser (AI-Detector Evading)",
+    description="Paraphrase full text with diversity and reordering control to reduce AI detection."
 )
 
 iface.launch()