Upload chatterbox_utils.py with huggingface_hub

chatterbox_utils.py

@@ -129,19 +129,18 @@ def load_chatterbox(device="cuda"):
     SESSIONS["tokenizer"] = AutoTokenizer.from_pretrained(MODEL_ID)
 
 def warmup_chatterbox():
-    """CPU-only download of model files for caching"""
-    print("🔥 Warming up Chatterbox (Downloading files)...")
+    """v92: Pre-download model files in background"""
+    print("📥 Caching Chatterbox weights (ONNX)...")
     try:
         AutoTokenizer.from_pretrained(MODEL_ID)
         hf_hub_download(repo_id=MODEL_ID, filename="default_voice.wav")
-        hf_hub_download(repo_id=MODEL_ID, filename="Cangjie5_TC.json")
         for sess_name in ["speech_encoder", "embed_tokens", "conditional_decoder", "language_model"]:
             fname = "onnx/" + (sess_name + ".onnx" if sess_name != "language_model" else "language_model.onnx")
             hf_hub_download(repo_id=MODEL_ID, filename=fname)
             hf_hub_download(repo_id=MODEL_ID, filename=fname + "_data")
-        print("✅ Chatterbox warmup complete")
+        print("✅ Chatterbox cached.")
     except Exception as e:
-        print(f"⚠️ Chatterbox warmup warning: {e}")
+        print(f"⚠️ Chatterbox cache warning: {e}")
 
 def run_chatterbox_inference(text, lang_id, speaker_wav_path=None):
     """Ported logic from model card with session reuse"""