Update app.py

app.py

@@ -1,75 +1,39 @@
-import os
-os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "0"
-
-from transformers import MBart50TokenizerFast, MBartForConditionalGeneration, AutoConfig
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 import gradio as gr
 
-# ---- Load model & tokenizer ----
-model_name = "Mudasir692/mbart-eng-ur"
-
-# Fix config issue
-config = AutoConfig.from_pretrained(model_name)
-if config.early_stopping is None:
-    config.early_stopping = True
-
-tokenizer = MBart50TokenizerFast.from_pretrained(model_name)
-model = MBartForConditionalGeneration.from_pretrained(model_name, config=config)
-
-# ---- Language mapping ----
-LANG_CODES = {
-    "Urdu": "ur_PK",
-    "Arabic": "ar_AR",
-    "Hindi": "hi_IN",
-}
+# Load model and tokenizer
+model_name = "Helsinki-NLP/opus-mt-en-ur"  # English ↔ Urdu model
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
 
-# ---- Translation function ----
-def translate_text(text, target_lang):
+def translate_text(text):
     if not text.strip():
-        return "Please enter some English text."
-
-    tgt_lang_code = LANG_CODES.get(target_lang, "ur_PK")
-    tokenizer.src_lang = "en_XX"
-    tokenizer.tgt_lang = tgt_lang_code
-
-    inputs = tokenizer(text, return_tensors="pt", padding=True)
-    translated_tokens = model.generate(**inputs)
-    output = tokenizer.decode(translated_tokens[0], skip_special_tokens=True)
-    return output
-
-# ---- Examples ----
-examples = [
-    ["How are you?", "Urdu"],
-    ["Where are you going?", "Arabic"],
-    ["This is my new project.", "Hindi"],
-]
-
-# ---- Gradio Interface ----
+        return "Please enter some text to translate."
+    
+    # Tokenize input
+    inputs = tokenizer(text, return_tensors="pt", truncation=True)
+
+    # Generate translation (explicit max_length fix)
+    translated_tokens = model.generate(
+        **inputs,
+        max_length=256,       # Fix for max_length=None error
+        num_beams=5,          # Enables beam search (more accurate)
+        early_stopping=True
+    )
+
+    # Decode output
+    translated_text = tokenizer.decode(translated_tokens[0], skip_special_tokens=True)
+    return translated_text
+
+# Build Gradio Interface
 app = gr.Interface(
     fn=translate_text,
-    inputs=[
-        gr.Textbox(label="Enter English Text", placeholder="Type your English sentence here...", lines=2),
-        gr.Dropdown(["Urdu", "Arabic", "Hindi"], label="Select Target Language", value="Urdu")
-    ],
-    outputs=gr.Textbox(label="Translation", lines=2),
-    examples=examples,
-    title="🌍 Multi-Language Translator",
-    description="""
-    <div style='text-align:center;'>
-        <h3>Translate English sentences into Urdu, Arabic, or Hindi using a fine-tuned mBART model.</h3>
-        <p style='color:gray;'>Built by <b>Khurram Basharat</b> — powered by Transformers & Gradio.</p>
-        <p><i>⏳ The model loads on first use, please wait a few seconds.</i></p>
-    </div>
-    """,
-    css="""
-        body {
-            background: linear-gradient(to bottom right, #f1f2f6, #dff9fb);
-            font-family: 'Segoe UI', sans-serif;
-        }
-        .gr-button-primary {
-            background-color: #1e3799 !important;
-            color: white !important;
-        }
-    """,
+    inputs=gr.Textbox(lines=3, placeholder="Enter English or Urdu text..."),
+    outputs="text",
+    title="English ↔ Urdu Translator",
+    description="Translate between English and Urdu using a Hugging Face translation model.",
 )
 
-app.launch()
+# Launch app
+if __name__ == "__main__":
+    app.launch(server_name="0.0.0.0", server_port=7860)