Struct LBFGS 

pub struct LBFGS { /* private fields */ }

L-BFGS optimizer.

Ported from pytorch. Heavily inspired by miniFunc

See also:

• L-BFGS

Note

This optimizer is memory intensive

Implementations

impl LBFGS

pub fn to_record(&self) -> OptimizerRecord

Decompose the optimizer state into a serializable OptimizerRecord (burnpack format).

L-BFGS keeps a single global state rather than per-parameter state, so its tensors are named directly (e.g. history_s.0) and carry no parameter id.

pub fn load_record(self, record: OptimizerRecord) -> Self

Load the optimizer state from an OptimizerRecord.

State tensors are materialized on the default device; the state is migrated to the gradient device on the next step, so no device argument is needed.

pub fn into_bytes(&self) -> Result<Bytes, RecordError>

Serialize the optimizer state to an in-memory burnpack byte buffer.

pub fn from_bytes(self, bytes: Bytes) -> Result<Self, RecordError>

Load the optimizer state from an in-memory burnpack byte buffer.

pub fn save<P: AsRef<Path>>(&self, path: P) -> Result<(), RecordError>

Available on crate feature std only.

Save the optimizer state to a burnpack file on disk.

pub fn load<P: AsRef<Path>>(self, path: P) -> Result<Self, RecordError>

Available on crate feature std only.

Load the optimizer state from a burnpack file on disk.

pub fn step<M, F>( &mut self, lr: LearningRate, module: M, closure: F, ) -> (M, f64)

where M: AutodiffModule + Clone, F: FnMut(M) -> (f64, GradientsParams),

A single optimization step for any tensor that represents the parameters of a model.

pub fn to_device(self, device: &Device) -> Self

Moves the optimizer state to the specified device.

Trait Implementations

impl Clone for LBFGS

fn clone(&self) -> LBFGS

Returns a duplicate of the value.

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

Performs copy-assignment from source.