Update modules/whisper/whisper_base.py

modules/whisper/whisper_base.py

@@ -671,16 +671,33 @@ class WhisperBase(ABC):
 
     @staticmethod
     def cache_parameters(
-        whisper_params: WhisperValues,
-        add_timestamp: bool
+        params: WhisperValues,
+        file_format: str = "SRT",
+        add_timestamp: bool = True
     ):
-        """cache parameters to the yaml file"""
+        """Cache parameters to the yaml file"""
         cached_params = load_yaml(DEFAULT_PARAMETERS_CONFIG_PATH)
-        cached_whisper_param = whisper_params.to_yaml()
-        cached_yaml = {**cached_params, **cached_whisper_param}
+        param_to_cache = params.to_dict()
+
+        cached_yaml = {**cached_params, **param_to_cache}
         cached_yaml["whisper"]["add_timestamp"] = add_timestamp
+        cached_yaml["whisper"]["file_format"] = file_format
+
+        suppress_token = cached_yaml["whisper"].get("suppress_tokens", None)
+        if suppress_token and isinstance(suppress_token, list):
+            cached_yaml["whisper"]["suppress_tokens"] = str(suppress_token)
+
+        if cached_yaml["whisper"].get("lang", None) is None:
+            cached_yaml["whisper"]["lang"] = AUTOMATIC_DETECTION.unwrap()
+        else:
+            language_dict = whisper.tokenizer.LANGUAGES
+            cached_yaml["whisper"]["lang"] = language_dict[cached_yaml["whisper"]["lang"]]
+
+        if cached_yaml["vad"].get("max_speech_duration_s", float('inf')) == float('inf'):
+            cached_yaml["vad"]["max_speech_duration_s"] = GRADIO_NONE_NUMBER_MAX
 
-        save_yaml(cached_yaml, DEFAULT_PARAMETERS_CONFIG_PATH)
+        if cached_yaml is not None and cached_yaml:
+            save_yaml(cached_yaml, DEFAULT_PARAMETERS_CONFIG_PATH)
 
     @staticmethod
     def resample_audio(audio: Union[str, np.ndarray],