| from transformers import PretrainedConfig | |
| class SpeechLLMModelConfig(PretrainedConfig): | |
| model_type = "custom_model" | |
| def __init__(self, **kwargs): | |
| super().__init__(**kwargs) | |
| self.audio_enc_dim = 1280 | |
| self.llm_dim = 2048 | |
| self.audio_processor_name = "facebook/hubert-large-ls960-ft" | |
| self.audio_encoder_name = 'facebook/hubert-xlarge-ll60k' | |
| self.llm_model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0" | |
| self.llm_model_checkpoint = "hf_repo/llm_model_checkpoint" | |