Skip to content

vllm_omni.model_executor.models.personaplex.configuration_helium

Configuration for the PersonaPlex Helium temporal transformer.

HeliumConfig

Bases: PretrainedConfig

Minimal HF config for the Moshi temporal LM backbone.

The defaults are measured from the PersonaPlex checkpoint rather than inferred from Moshi's higher-level kwargs.

attention_bias instance-attribute 

attention_bias = attention_bias

head_dim instance-attribute 

head_dim = head_dim

hidden_act instance-attribute 

hidden_act = hidden_act

hidden_size instance-attribute 

hidden_size = hidden_size

intermediate_size instance-attribute 

intermediate_size = intermediate_size

keys_to_ignore_at_inference class-attribute instance-attribute 

keys_to_ignore_at_inference = ('past_key_values',)

max_position_embeddings instance-attribute 

max_position_embeddings = max_position_embeddings

mlp_bias instance-attribute 

mlp_bias = mlp_bias

model_type class-attribute instance-attribute 

model_type = 'helium'

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

rms_norm_eps instance-attribute 

rms_norm_eps = rms_norm_eps

rope_theta instance-attribute 

rope_theta = rope_theta

sliding_window instance-attribute 

sliding_window = sliding_window

use_cache instance-attribute 

use_cache = use_cache

vocab_size instance-attribute 

vocab_size = vocab_size