Upload app.py

app.py

@@ -12,7 +12,7 @@ DEFAULT_MAX_NEW_TOKENS = 1024
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 ACCESS_TOKEN = os.getenv("HF_TOKEN", "")
 
-model_id = "Qwen/Qwen2.5-14B-Instruct"
+model_id = "microsoft/Phi-3-small-8k-instruct"
 model = AutoModelForCausalLM.from_pretrained(
     model_id,
     torch_dtype=torch.float16,