boostr 0.1.0

ML framework built on numr - attention, quantization, model architectures
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

pub mod adagrad;
pub mod adamw;
pub mod grad_accumulator;
pub mod grad_clip;
pub mod grad_scaler;
pub mod lamb;
pub mod lr_schedule;
pub mod sgd;
pub mod traits;

pub use adagrad::{AdaGrad, AdaGradConfig};
pub use adamw::{AdamW, AdamWConfig};
pub use grad_accumulator::GradAccumulator;
pub use grad_clip::{clip_grad_norm, clip_grad_norm_per_param, clip_grad_value};
pub use grad_scaler::GradScaler;
pub use lamb::{Lamb, LambConfig};
pub use lr_schedule::{DecayShape, LrSchedule};
pub use sgd::{Sgd, SgdConfig};
pub use traits::Optimizer;