config.py

from transformers import PretrainedConfig

class SpeechLLMModelConfig(PretrainedConfig):
    model_type = "custom_model"
    
    def __init__(self, audio_enc_dim=1280, llm_dim=2048, **kwargs):
        super().__init__(**kwargs)
        self.audio_enc_dim = audio_enc_dim
        self.llm_dim = llm_dim