Update app.py

app.py

@@ -70,6 +70,7 @@ tokenizer = AutoTokenizer.from_pretrained(
     , token= token,)
 
 
+accelerator = Accelerator()
 
 model = AutoModelForCausalLM.from_pretrained(model_id, token= token, 
                                                  # torch_dtype= torch.uint8, 
@@ -78,12 +79,14 @@ model = AutoModelForCausalLM.from_pretrained(model_id, token= token,
                                              attn_implementation="flash_attention_2",
                                              low_cpu_mem_usage=True,
                                             
-                                             device_map='cuda',
+                                             # device_map='cuda',
+                                             device_map=accelerator.device_map,
                                              
                                             )
 
 
 # 
+model = accelerator.prepare(model)
 
 
 # device_map = infer_auto_device_map(model, max_memory={0: "79GB", "cpu":"65GB" })
@@ -104,7 +107,7 @@ def respond(
     top_p,
 ):
     messages = [{"role": "user", "content": "Hello, how are you?"}]
-    input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt").to('cuda')
+    input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt").to(accelerator.device) #.to('cuda')
 ## <BOS_TOKEN><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>
     # with autocast():
     gen_tokens = model.generate(