Add download file from hf model

app.py

@@ -15,10 +15,11 @@ from utils import load_tokenizer
 
 from llama_index.llms.gemini import Gemini
 from llama_index.core.tools import RetrieverTool
-
+from huggingface_hub import hf_hub_download
 
 GEMINI_API_KEY = os.getenv(key="GEMINI_API_KEY")
 QDRANT_API_KEY = os.getenv(key="QDRANT_API_KEY")
+HF_TOKEN_KEY = os.getenv(key="HF_TOKEN_KEY")
 device = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
 
 async def initialize_model() -> Dict:
@@ -27,7 +28,22 @@ async def initialize_model() -> Dict:
     Returns:
         model_dict: Dict: Dictionary stores neccessary models
     """
-    
+    if not os.path.exists("./pretrained/colpaligemma-3b-mix-448-base"):
+        os.makedirs("./pretrained/colpaligemma-3b-mix-448-base", exist_ok=True)
+        files_to_download = ["adapter_model.safetensors",
+                            "config.json",
+                            "model-00002-of-00002.safetensors",
+                            "model-00002-of-00002.safetensors",
+                            "preprocessor_config.json",
+                            "tokenizer.json",
+                            "tokenizer.model",
+                            "tokenizer_config.json"]
+        for file in files_to_download:
+            hf_hub_download(repo_id="dnnhhuy/colpaligemma-3b-mix-448-base",
+                            filename=file,
+                            token=HF_TOKEN_KEY,
+                            local_dir="./pretrained/colpaligemma-3b-mix-448-base")
+        
     model = ColPali.from_pretrained(model_dir='./pretrained/colpaligemma-3b-mix-448-base', torch_dtype=torch.bfloat16)
     tokenizer = load_tokenizer(tokenizer_dir='./pretrained/colpaligemma-3b-mix-448-base')
     processor = ColPaliProcessor(tokenizer=tokenizer).from_pretrained(pretrained_dir='./pretrained/colpaligemma-3b-mix-448-base')