Update app.py

app.py

@@ -4,11 +4,10 @@ from threading import Thread
 from transformers import StoppingCriteria, StoppingCriteriaList
 import torch
 import os 
-from unsloth import FastLanguageModel
 model_name = "microsoft/Phi-3-medium-128k-instruct"
 from transformers import AutoModelForCausalLM, AutoTokenizer
-model, tokenizer = FastLanguageModel.from_pretrained(model_name, device_map='cuda')
-
+model = AutoModelForCausalLM.from_pretrained(model_name, device_map='cuda')
+tokenizer = AutoTokenizer.from_pretrained(model_name)
 
 class StopOnTokens(StoppingCriteria):
     def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool: