Update app.py

app.py

@@ -1,24 +1,29 @@
 # Install necessary libraries
 # !pip install transformers
 
-from transformers import AutoTokenizer, pipeline
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 import torch
 
 # Model and prompt details
 model_name = "mlabonne/llama-2-7b-guanaco"
 prompt = "What is a large language model?"
 
-# Load tokenizer and pipeline
+# Load tokenizer and model
 tokenizer = AutoTokenizer.from_pretrained(model_name)
-text_generation_pipeline = pipeline(
-    "text-generation",
-    model=model_name,
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
     torch_dtype=torch.float16,
     device_map="auto",
+    offload_folder="path/to/offload/folder"  # Replace with the path to the offload folder
 )
 
 # Generate text using the provided prompt
-sequences = text_generation_pipeline(
+sequences = pipeline(
+    "text-generation",
+    model=model,
+    tokenizer=tokenizer,
+    device=0,  # Change to the appropriate device index or "cuda" if using GPU
+)(
     f'<s>[INST] {prompt} [/INST]',
     do_sample=True,
     top_k=10,