submodule

app.py

@@ -1,21 +1,10 @@
-import os
-
 import gradio as gr
 import spaces
 from guidance import gen
 from guidance.models import Transformers
-from transformers import AutoModelForCausalLM, AutoTokenizer
+from mymodule.utils import load_model_and_tok
 
-AUTH_TOKEN = os.environ.get("HF_TOKEN", False)
-model = AutoModelForCausalLM.from_pretrained(
-    "meta-llama/Llama-3.2-3B-Instruct",
-    token=AUTH_TOKEN,
-)
-tok = AutoTokenizer.from_pretrained(
-    "meta-llama/Llama-3.2-3B-Instruct",
-    token=AUTH_TOKEN,
-)
-model = model.to("cuda").eval()
+model, tok = load_model_and_tok()
 
 
 @spaces.GPU(duration=10)

mymodule/utils.py

@@ -0,0 +1,17 @@
+import os
+
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+
+def load_model_and_tok():
+    AUTH_TOKEN = os.environ.get("HF_TOKEN", False)
+    model = AutoModelForCausalLM.from_pretrained(
+        "meta-llama/Llama-3.2-3B-Instruct",
+        token=AUTH_TOKEN,
+    )
+    tok = AutoTokenizer.from_pretrained(
+        "meta-llama/Llama-3.2-3B-Instruct",
+        token=AUTH_TOKEN,
+    )
+    model = model.to("cuda").eval()
+    return model, tok