Update custom model files, README, and requirements

asr_modeling.py

@@ -38,7 +38,7 @@ class ASRModel(PreTrainedModel, GenerationMixin):
     _is_loading_from_pretrained: bool = False
     _pretrained_model_path: Optional[str] = None
 
-    TRANSCRIBE_PROMPT = "Transcribe speech to text"
+    TRANSCRIBE_PROMPT = ""
 
     @classmethod
     def from_pretrained(cls, pretrained_model_name_or_path: str, *args, **kwargs) -> "ASRModel":

asr_processing.py

@@ -17,7 +17,7 @@ class ASRProcessor(ProcessorMixin):
     feature_extractor_class = "AutoFeatureExtractor"
     tokenizer_class = "AutoTokenizer"
     AUDIO_TOKEN = "<audio>"
-    TRANSCRIBE_PROMPT = "Transcribe speech to text"
+    TRANSCRIBE_PROMPT = ""
     # Default conv layers for Whisper/GLM-ASR: [(pad, kernel, stride), ...]
     DEFAULT_ENCODER_CONV_LAYERS = [(1, 3, 1), (1, 3, 2)]