{ "_name_or_path": "./hf_hub/hcx", "architectures": [ "OmniSpeechHyperCLOVAXForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 100257, "deepspeed_config": "./scripts/zero2.json", "end_token_id": 100257, "eos_token_id": 100257, "fullduplex_dropout": 0.1, "fullduplex_nhead": 8, "fullduplex_num_classes": 3, "head_dim": 128, "hidden_act": "silu", "hidden_size": 1024, "initializer_range": 0.02, "intermediate_size": 4096, "max_position_embeddings": 8192, "mlp_bias": false, "mm_tunable_parts": "speech_projector,backbone", "model_type": "omni_speech_HyperCLOVAX", "num_attention_heads": 16, "num_hidden_layers": 24, "num_key_value_heads": 8, "pad_token_id": 100257, "pretraining_tp": 1, "resid_pdrop": 0.2, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 500000, "speech_encoder": "./models/speech_encoder/whisper-large-v3", "speech_encoder_ds_rate": 5, "speech_encoder_hidden_size": 1280, "speech_encoder_type": "whisper", "speech_projector_lr": null, "speech_projector_type": "linear", "tie_word_embeddings": true, "tokenizer_model_max_length": 2048, "tokenizer_padding_side": "right", "torch_dtype": "float16", "transformers_version": "4.45.1", "use_cache": true, "use_duplex": false, "vocab_size": 110592 }