Update app.py

app.py

@@ -48,6 +48,36 @@ LANGUAGES = {
     "Bodo": "brx_Deva"
 }
 
+@spaces.GPU
+def generate_for_examples(
+    tgt_lang: str,
+    message: str,
+    max_new_tokens: int = 1024,
+    temperature: float = 0.6,
+    top_p: float = 0.9,
+    top_k: int = 50,
+    repetition_penalty: float = 1.2,
+) -> str:
+    conversation = []
+    conversation.append({"role": "user", "content": f"Translate the following text to {tgt_lang}: {message}"})
+
+    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
+    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
+        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
+    input_ids = input_ids.to(model.device)
+
+    outputs = model.generate(
+        input_ids=input_ids,
+        max_new_tokens=max_new_tokens,
+        do_sample=True,
+        top_p=top_p,
+        top_k=top_k,
+        temperature=temperature,
+        num_beams=1,
+        repetition_penalty=repetition_penalty,
+    )
+    
+    return tokenizer.decode(outputs[0][input_ids.shape[1]:], skip_special_tokens=True)
 
 
 @spaces.GPU
@@ -168,7 +198,7 @@ with gr.Blocks(theme=gr.themes.Default(), css=css) as demo:
                 text_input
             ], 
             outputs=text_output,
-            fn=generate,
+            fn=generate_for_examples,
             cache_examples=True,
             examples_per_page=5
         )