pub struct GdnCfg {
pub num_v_heads: usize,
pub num_k_heads: usize,
pub key_head_dim: usize,
pub value_head_dim: usize,
pub conv_kernel: usize,
pub hidden_size: usize,
pub rms_eps: f64,
pub output_gate_sigmoid: bool,
}Fields§
§num_v_heads: usize§num_k_heads: usize§key_head_dim: usize§value_head_dim: usize§conv_kernel: usize§rms_eps: f64§output_gate_sigmoid: boolQwen4-exp trains the output gate as sigmoid(z); older GDN families use SiLU(z). This is part of the operator, not a sampling option.
Implementations§
Trait Implementations§
Auto Trait Implementations§
impl Freeze for GdnCfg
impl RefUnwindSafe for GdnCfg
impl Send for GdnCfg
impl Sync for GdnCfg
impl Unpin for GdnCfg
impl UnsafeUnpin for GdnCfg
impl UnwindSafe for GdnCfg
Blanket Implementations§
Source§impl<T> BorrowMut<T> for Twhere
T: ?Sized,
impl<T> BorrowMut<T> for Twhere
T: ?Sized,
Source§fn borrow_mut(&mut self) -> &mut T
fn borrow_mut(&mut self) -> &mut T
Mutably borrows from an owned value. Read more