nemo_automodel.components.models.deepseek_v32.config

class nemo_automodel.components.models.deepseek_v32.config.DeepseekV32Config(
    vocab_size: int = 129280,
    hidden_size: int = 7168,
    intermediate_size: int = 18432,
    moe_intermediate_size: int = 2048,
    num_hidden_layers: int = 61,
    num_attention_heads: int = 128,
    num_key_value_heads: int = 128,
    n_shared_experts: int = 1,
    n_routed_experts: int = 256,
    ep_size: int = 1,
    routed_scaling_factor: float = 2.5,
    kv_lora_rank: int = 512,
    q_lora_rank: int = 1536,
    qk_head_dim: int = 192,
    qk_nope_head_dim: int = 128,
    qk_rope_head_dim: int = 64,
    v_head_dim: int = 128,
    hidden_act: str = 'silu',
    max_position_embeddings: int = 163840,
    initializer_range: float = 0.02,
    rms_norm_eps: float = 1e-06,
    use_cache: bool = True,
    pad_token_id: int | None = None,
    bos_token_id: int = 0,
    eos_token_id: int = 1,
    pretraining_tp: int = 1,
    tie_word_embeddings: bool = False,
    rope_theta: float = 10000.0,
    rope_scaling: dict | None = None,
    attention_bias: bool = False,
    attention_dropout: float = 0.0,
    num_experts_per_tok: int = 8,
    n_group: int = 8,
    topk_group: int = 4,
    first_k_dense_replace: int = 3,
    norm_topk_prob: bool = True,
    scoring_func: str = 'sigmoid',
    aux_loss_alpha: float = 0.001,
    seq_aux: bool = True,
    index_n_heads: int = 64,
    index_head_dim: int = 128,
    index_topk: int = 2048,
    torch_dtype: str = 'bfloat16',
    kwargs = {}
)