Update app.py

app.py

@@ -27,5 +27,28 @@ inputs = tokenizer(
 )
 input_ids = inputs["input_ids"]#.to("cuda")
 
+generation_config = GenerationConfig(
+    do_sample=True,
+    temperature=0.8,  # default 0.1
+    top_p=0.995,      # default 0.75
+    top_k=100,        # default 80
+    repetition_penalty=1.5,
+    max_new_tokens=2,
+)
+
 if instruction:
-    st.write(tokenizer.decode(input_ids[0]))
+    with torch.no_grad():
+    outputs = model.generate(
+        input_ids=input_ids,
+        attention_mask=torch.ones_like(input_ids),
+        generation_config=generation_config,
+        return_dict_in_generate=True, 
+        output_scores=True
+    )
+    
+    output_text = tokenizer.decode(
+        outputs['sequences'][0].cuda(), 
+        skip_special_tokens=False
+    ).strip()
+    
+    st.write(output_text)