y-ren16/MCLP-RPTTS
248
1{2 "architectures": [3 "StepAudio2ForCausalLM"4 ],5 "auto_map": {6 "AutoConfig": "configuration_step_audio_2.StepAudio2Config",7 "AutoModelForCausalLM": "modeling_step_audio_2.StepAudio2ForCausalLM"8 },9 "model_type": "step_audio_2",10 "text_config": {11 "hidden_size": 3584,12 "intermediate_size": 18944,13 "num_attention_heads": 28,14 "num_attention_groups": 4,15 "num_key_value_heads": 4,16 "num_hidden_layers": 28,17 "max_seq_len": 16384,18 "vocab_size": 158720,19 "rms_norm_eps": 1e-06,20 "eos_token_id": 151643,21 "pad_token_id": 151643,22 "rope_theta": 1000000.0,23 "max_position_embeddings": 16384,24 "rope_scaling": null,25 "torch_dtype": "bfloat16"26 },27 "audio_encoder_config": {28 "n_mels": 128,29 "n_audio_ctx": 1500,30 "n_audio_state": 1280,31 "n_audio_head": 20,32 "n_audio_layer": 32,33 "n_codebook_size": 4096,34 "llm_dim": 3584,35 "kernel_size": 3,36 "adapter_stride": 237 }38}39 