# config/config.py import os from groq import Groq # Groq API 設定 GROQ_API_KEY = os.getenv('GROQ_API_KEY') MODEL = "llama-3.1-70b-versatile" AUDIO_MODEL = "whisper-large-v3" # 檔案路徑設定 BASE_PATH = "public" AUDIO_PATH = os.path.join(BASE_PATH, "audios") IMAGE_PATH = os.path.join(BASE_PATH, "images") # 模型設定 SPEECH_MODEL_CONFIG = { "tts_model": "microsoft/speecht5_tts", "vocoder_model": "microsoft/speecht5_hifigan", "speaker_dataset": "Matthijs/cmu-arctic-xvectors", "max_token_limit": 600 } LLAVA_MODEL_CONFIG = { "model_path": "llava-hf/llava-1.5-7b-hf", "load_in_4bit": True, "max_new_tokens": 500 }