Files
aha/src/models/minicpm4/config.rs
T

32 lines
926 B
Rust
Raw Normal View History

2025-09-25 12:09:25 +08:00
use candle_nn::Activation;
#[derive(Debug, Clone, PartialEq, serde::Deserialize)]
pub struct RopeScalingConfig {
pub rope_type: String,
pub long_factor: Vec<f32>,
pub short_factor: Vec<f32>,
pub original_max_position_embeddings: usize,
}
#[derive(Debug, Clone, PartialEq, serde::Deserialize)]
pub struct MiniCPM4Config {
pub bos_token_id: u32,
pub eos_token_id: Vec<u32>,
pub hidden_act: Activation,
pub hidden_size: usize,
pub intermediate_size: usize,
pub max_position_embeddings: usize,
pub num_attention_heads: usize,
pub num_hidden_layers: usize,
pub num_key_value_heads: usize,
pub rms_norm_eps: f64,
pub rope_scaling: RopeScalingConfig,
pub torch_dtype: String,
pub vocab_size: usize,
// pub use_mup: bool,
pub scale_emb:f32,
pub dim_model_base: usize,
pub scale_depth: f32,
// pub rope_theta: f32,
// pub kv_channels: i32,
}