Bigvox-Exaone4-Audio / config.json
bigdefence's picture
Update config.json
d2961f4 verified
{
"architectures": [
"OmniSpeechExaoneForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 1,
"deepspeed_config": "./scripts/zero2.json",
"eos_token_id": 361,
"fullduplex_dropout": 0.1,
"fullduplex_nhead": 8,
"fullduplex_num_classes": 3,
"head_dim": 64,
"hidden_act": "silu",
"hidden_size": 2048,
"initializer_range": 0.02,
"intermediate_size": 4096,
"layer_types": [
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention"
],
"max_position_embeddings": 65536,
"mlp_bias": false,
"mm_tunable_parts": "speech_projector,backbone",
"model_type": "omni_speech_exaone",
"num_attention_heads": 32,
"num_hidden_layers": 30,
"num_key_value_heads": 8,
"pad_token_id": 0,
"pretraining_tp": 1,
"rms_norm_eps": 1e-05,
"rope_scaling": {
"factor": 16.0,
"high_freq_factor": 4.0,
"low_freq_factor": 1.0,
"original_max_position_embeddings": 8192,
"rope_type": "llama3"
},
"rope_theta": 1000000.0,
"sliding_window": null,
"sliding_window_pattern": null,
"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": "4.54.0",
"use_cache": true,
"use_duplex": false,
"vocab_size": 102400
}