remove device as loaded with accelerate

app.py

@@ -45,7 +45,7 @@ last_messages = []
 def load_model():
     ankerbot_model = AutoModelForCausalLM.from_pretrained(
         model_name,
-        device_map="auto",
+        device_map="cuda:0",
         torch_dtype=torch.float16,
         use_cache=True,
         offload_folder="../offload"
@@ -55,8 +55,8 @@ def load_model():
         torch_dtype=torch.float16,
         truncation=True,
         padding=True, )
-    generator = pipeline(task="text-generation", model=ankerbot_model, tokenizer=ankerbot_tokenizer, torch_dtype=torch.float16, trust_remote_code=False, device="cuda:0") # True for flash-attn2 else False
-    generator_mini = pipeline(task="text-generation", model=ankerbot_model, tokenizer=ankerbot_tokenizer, torch_dtype=torch.float16, trust_remote_code=False, device="cuda:0") # True for flash-attn2 else False
+    generator = pipeline(task="text-generation", model=ankerbot_model, tokenizer=ankerbot_tokenizer, torch_dtype=torch.float16, trust_remote_code=False) # True for flash-attn2 else False
+    generator_mini = pipeline(task="text-generation", model=ankerbot_model, tokenizer=ankerbot_tokenizer, torch_dtype=torch.float16, trust_remote_code=False) # True for flash-attn2 else False
     return (generator, generator_mini)
     
 _model_cache = None