Skip to content

vllm_omni.transformers_utils.configs.glm_tts

GLM-TTS config registration with transformers AutoConfig.

Registers GLMTTSConfig (model_type="glm_tts") so that AutoConfig.from_pretrained("path/to/glm-tts") returns the correct config class.

Note: GLM-TTS uses a Llama backbone, but we register a custom config to handle the special token IDs and flow model parameters.

GLMTTSConfig

Bases: PretrainedConfig

Llama-based AR model for text-to-speech token generation.

Special token IDs are loaded dynamically from the tokenizer at init time.

audio_token_end instance-attribute 

audio_token_end = audio_token_end

audio_token_start instance-attribute 

audio_token_start = audio_token_start

boa_token_id instance-attribute 

boa_token_id = boa_token_id

eoa_token_id instance-attribute 

eoa_token_id = eoa_token_id

hidden_act instance-attribute 

hidden_act = hidden_act

hidden_size instance-attribute 

hidden_size = hidden_size

input_frame_rate instance-attribute 

input_frame_rate = input_frame_rate

intermediate_size instance-attribute 

intermediate_size = intermediate_size

max_position_embeddings instance-attribute 

max_position_embeddings = max_position_embeddings

max_token_text_ratio instance-attribute 

max_token_text_ratio = max_token_text_ratio

mel_dim instance-attribute 

mel_dim = mel_dim

mel_framerate instance-attribute 

mel_framerate = mel_framerate

min_token_text_ratio instance-attribute 

min_token_text_ratio = min_token_text_ratio

model_type class-attribute instance-attribute 

model_type: str = 'glm_tts'

num_attention_heads instance-attribute 

num_attention_heads = num_attention_heads

num_hidden_layers instance-attribute 

num_hidden_layers = num_hidden_layers

num_key_value_heads instance-attribute 

num_key_value_heads = num_key_value_heads

ras_tau_r instance-attribute 

ras_tau_r = ras_tau_r

ras_top_k instance-attribute 

ras_top_k = ras_top_k

ras_top_p instance-attribute 

ras_top_p = ras_top_p

ras_win_size instance-attribute 

ras_win_size = ras_win_size

rms_norm_eps instance-attribute 

rms_norm_eps = rms_norm_eps

rope_theta instance-attribute 

rope_theta = rope_theta

sample_method instance-attribute 

sample_method = sample_method

speech_token_dim instance-attribute 

speech_token_dim = speech_token_dim

speech_token_vocab_size instance-attribute 

speech_token_vocab_size = speech_token_vocab_size

spk_embed_dim instance-attribute 

spk_embed_dim = spk_embed_dim

vocab_size instance-attribute 

vocab_size = vocab_size