removed disk_offload function

app.py

@@ -54,14 +54,14 @@ model_config = transformers.AutoConfig.from_pretrained(
 )
 
 # Downloading and Initializing the model
-model = disk_offload(transformers.AutoModelForCausalLM.from_pretrained(
+model = transformers.AutoModelForCausalLM.from_pretrained(
     model_id,
     trust_remote_code=True,
     config=model_config,
     # quantization_config=bnb_config,
     device_map='auto',
     token=hf_auth
-))
+)
 
 # enable evaluation mode to allow model inference
 model.eval()