Update app.py

app.py

@@ -11,7 +11,7 @@ bnb_config = BitsAndBytesConfig(
             load_in_4bit=True,
             bnb_4bit_use_double_quant=True,
             bnb_4bit_quant_type="nf4",
-            bnb_4bit_compute_dtype=torch.bfloat16
+            bnb_4bit_compute_dtype=torch.float16
         )
 
 # Load models and tokenizer efficiently
@@ -24,7 +24,7 @@ model = PeftModel.from_pretrained(model, model_id)
 
 def greet(text):
     with torch.no_grad():  # Disable gradient calculation for inference
-        batch = tokenizer(f'### Human: {text}', return_tensors='pt')  # Move tensors to device
+        batch = tokenizer(f'### Human: {text}\n### Assistant:', return_tensors='pt')  # Move tensors to device
         with torch.cuda.amp.autocast():  # Enable mixed-precision if available
             output_tokens = model.generate(**batch, max_new_tokens=25)
     return tokenizer.decode(output_tokens[0], skip_special_tokens=True)