Struct TrainConfig 

pub struct TrainConfig {
    pub max_grad_norm: Option<f32>,
    pub log_interval: usize,
    pub save_interval: Option<usize>,
    pub checkpoint_dir: Option<PathBuf>,
    pub mixed_precision: bool,
    pub gradient_accumulation_steps: usize,
}

Training configuration

Fields

max_grad_norm: Option<f32>

Maximum gradient norm for clipping (None = no clipping)

log_interval: usize

Print training progress every N steps

save_interval: Option<usize>

Save checkpoint every N epochs

checkpoint_dir: Option<PathBuf>

Directory to save checkpoints

mixed_precision: bool

Use mixed precision training

gradient_accumulation_steps: usize

Gradient accumulation steps (1 = no accumulation)

Simulates larger batch sizes by accumulating gradients over multiple mini-batches before performing an optimizer step. Effective batch size = batch_size * gradient_accumulation_steps

Implementations

impl TrainConfig

pub fn new() -> Self

Create a new training configuration

pub fn with_grad_clip(self, max_norm: f32) -> Self

Set gradient clipping norm

pub fn without_grad_clip(self) -> Self

Disable gradient clipping

pub fn with_log_interval(self, interval: usize) -> Self

Set logging interval

pub fn with_checkpoints(self, interval: usize, dir: PathBuf) -> Self

Set checkpoint saving

pub fn with_gradient_accumulation(self, steps: usize) -> Self

Set gradient accumulation steps

Simulates larger batch sizes by accumulating gradients over multiple mini-batches before performing an optimizer step. Effective batch size = batch_size * gradient_accumulation_steps

Trait Implementations

impl Clone for TrainConfig

fn clone(&self) -> TrainConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for TrainConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for TrainConfig

fn default() -> Self

Returns the “default value” for a type.