Set the environment variable MODEL_PATH at entrypoint

app/utils/download_model.py

@@ -2,9 +2,23 @@ import os
 import requests
 from transformers import AutoModel
 
+def model_file_exists_and_valid(model_file_path):
+    # Check if the model file exists and has a size greater than 0
+    return os.path.exists(model_file_path) and os.path.getsize(model_file_path) > 0
+
 def download_hf_model():
+    '''
+    Model File Path for HF Models: The download_hf_model function now includes a default model file path (pytorch_model.bin) check. 
+    Adjust this path based on the expected model file type (e.g., TensorFlow or Flax models might have different names).
+    '''
     model_name = os.getenv("HF_MODEL_NAME")
     model_dir = f"/home/user/data/models/{model_name}"
+    model_file_path = os.path.join(model_dir, "pytorch_model.bin") # Assuming PyTorch model for simplicity
+    
+    if model_file_exists_and_valid(model_file_path):
+        print(f"Model {model_name} already downloaded.")
+        os.environ['MODEL_PATH'] = model_file_path
+        return
     
     # Authenticate with Hugging Face using the token, if available
     hf_token = os.getenv("HF_TOKEN")
@@ -16,6 +30,7 @@ def download_hf_model():
     model = AutoModel.from_pretrained(model_name)
     model.save_pretrained(model_dir)
     print(f"Model {model_name} downloaded and saved to {model_dir}")
+    os.environ['MODEL_PATH'] = model_file_path
 
 def download_gguf_model():
     model_name = os.getenv("HF_MODEL_NAME")
@@ -23,9 +38,13 @@ def download_gguf_model():
     os.makedirs(model_dir, exist_ok=True)
     
     model_url = os.getenv("GGUF_MODEL_URL")  # Assuming URL is provided as an env variable
-
     model_file_path = os.path.join(model_dir, os.path.basename(model_url))
     
+    if model_file_exists_and_valid(model_file_path):
+        print(f"Model {model_name} already downloaded.")
+        os.environ['MODEL_PATH'] = model_file_path
+        return
+    
     print(f"Downloading model from {model_url}...")
     response = requests.get(model_url, stream=True)
     if response.status_code == 200:
@@ -34,6 +53,7 @@ def download_gguf_model():
         print(f"Model downloaded and saved to {model_file_path}")
     else:
         print(f"Failed to download the model. Status code: {response.status_code}")
+    os.environ['MODEL_PATH'] = model_file_path
 
 def download_model():
     model_class = os.getenv("MODEL_CLASS")