VocalNet-qwen25-7B / config.json
mumuye's picture
Upload folder using huggingface_hub
faec628 verified
{
"_name_or_path": "./checkpoints/qwen25-7b-instruct-ut-va-mg-be-3epoch-s2s",
"architectures": [
"OmniSpeech2SQwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"decoder_dropout": 0.1,
"decoder_ffn_dim": 14336,
"decoder_hidden_size": 3584,
"decoder_num_heads": 32,
"decoder_num_layers": 4,
"deepspeed_config": null,
"encoder_num_layers": 2,
"eos_token_id": 151645,
"fullduplex_dropout": 0.1,
"fullduplex_nhead": 8,
"fullduplex_num_classes": 3,
"gen_loss_weight": 1.0,
"group_size": 3,
"hidden_act": "silu",
"hidden_size": 3584,
"initializer_range": 0.02,
"intermediate_size": 18944,
"llm_hidden_size": 3584,
"max_position_embeddings": 32768,
"max_seq_length": 8192,
"max_speech_tokens": 4096,
"max_window_layers": 28,
"mm_tunable_parts": "speech_generator",
"model_type": "omni_speech2s_qwen",
"mtp_num": 5,
"num_attention_heads": 28,
"num_hidden_layers": 28,
"num_key_value_heads": 4,
"rms_norm_eps": 1e-06,
"rope_theta": 1000000.0,
"sliding_window": null,
"special_tokens": 4,
"speech_bos_token_id": 6561,
"speech_encoder": "./models/speech_encoder/whisper-large-v3",
"speech_encoder_ds_rate": 5,
"speech_encoder_hidden_size": 1280,
"speech_encoder_type": "whisper",
"speech_eos_token_id": 6563,
"speech_generator_config": "./scripts/mtp/qwen_ar_config_5.yaml",
"speech_generator_type": "ar_mtp",
"speech_max_position_embeddings": 2048,
"speech_padding_token_id": 6564,
"speech_projector_lr": null,
"speech_projector_type": "linear",
"speech_sos_token_id": 6562,
"speech_token_num": 15,
"switch_token_id": 6565,
"tie_word_embeddings": false,
"tokenizer_model_max_length": 4096,
"tokenizer_padding_side": "right",
"torch_dtype": "bfloat16",
"transformers_version": "4.43.4",
"txt_token_num": 5,
"unit_vocab_size": 6561,
"use_cache": true,
"use_duplex": false,
"use_sliding_window": false,
"vocab_size": 152064
}