Lower limit

app.py

@@ -16,12 +16,12 @@ quant_config = BitsAndBytesConfig(
 )
 
 model = MistralForCausalLM.from_pretrained("mistralai/Mistral-7B-v0.1", 
-    quantization_config=quant_config,).to("cuda")
+    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)
 
-@spaces.GPU(duration=200)
+@spaces.GPU(duration=20)
 def translate(input_text):
     # iteratively generate
     input_text = input_text.strip()