add change models loading

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 import os
 import ctranslate2
 from transformers import M2M100Tokenizer
+from huggingface_hub import snapshot_download
 import time
 
 # Model settings
@@ -12,9 +13,15 @@ TOKENIZER_PATH = os.getenv("TOKENIZER_PATH", "skypro1111/m2m100-ukr-verbalizatio
 def init_model(model_path=MODEL_PATH, tokenizer_path=TOKENIZER_PATH):
     print("\nInitializing CTranslate2 model and tokenizer...")
     
+    # Download the model from HuggingFace Hub
+    local_model_path = snapshot_download(
+        repo_id=model_path,
+        allow_patterns=["*.bin", "*.json", "tokenizer.json", "vocab.json"],
+    )
+    
     # Initialize translator with optimizations for CPU
     translator = ctranslate2.Translator(
-        model_path,
+        local_model_path,
         device="cpu",
         compute_type="int8",  # Use INT8 quantization for CPU
         intra_threads=4,      # Adjust based on CPU cores available