Use cuda

app.py

@@ -16,8 +16,8 @@ quant_config = BitsAndBytesConfig(
 )
 
 model = MistralForCausalLM.from_pretrained("mistralai/Mistral-7B-v0.1", 
-    quantization_config=quant_config,
-    device_map="auto",)
+    quantization_config=quant_config,).to("cuda")
+    #device_map="auto",)
 model = PeftModel.from_pretrained(model, "lang-uk/dragoman")
 tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-v0.1", use_fast=False, add_bos_token=False)