Update app.py

app.py

@@ -4,11 +4,11 @@ import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 from threading import Thread
 
-torch.set_default_device("cuda")
 tokenizer = AutoTokenizer.from_pretrained(
     "cognitivecomputations/dolphin-2.9.1-mixtral-1x22b",
     trust_remote_code=True
 )
+
 model = AutoModelForCausalLM.from_pretrained(
     "cognitivecomputations/dolphin-2.9.1-mixtral-1x22b",
     torch_dtype="auto",
@@ -23,7 +23,7 @@ system_prompt = "<|im_start|>system\nYou are Dolphin, a helpful AI assistant.<|i
 def predict(message, history):
     history_transformer_format = history + [[message, ""]]
     messages = system_prompt + "".join(["".join(["\n<|im_start|>user\n" + item[0], "<|im_end|>\n<|im_start|>assistant\n" + item[1]]) for item in history_transformer_format])
-    input_ids = tokenizer([messages], return_tensors="pt").to('cuda')
+    input_ids = tokenizer([messages], return_tensors="pt")
     streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
     generate_kwargs = dict(
         input_ids,