Update app.py

app.py

@@ -6,9 +6,9 @@ from transformers import StoppingCriteria, StoppingCriteriaList, TextIteratorStr
 from threading import Thread
 
 # Loading the tokenizer and model from Hugging Face's model hub.
-tokenizer = AutoTokenizer.from_pretrained("0x7194633/fialka-13B-v3.1")
-model = AutoModelForCausalLM.from_pretrained("0x7194633/fialka-13B-v3.1", load_in_8bit=True)
-model.to("cuda:0")
+if torch.cuda.is_available():
+    tokenizer = AutoTokenizer.from_pretrained("0x7194633/fialka-13B-v3.1")
+    model = AutoModelForCausalLM.from_pretrained("0x7194633/fialka-13B-v3.1", load_in_8bit=True, device_map="auto")
 
 
 # Defining a custom stopping criteria class for the model's text generation.