Realm

oa::ml::training::PpoUpdateProgram

Reusable fixed-geometry PPO backward and AdamW programs. Retain this value across trainers when checkpointing between rollouts. Each minibatch reuses one bounded tape’s storage, accumulating sample-weighted gradients before exactly one AdamW step per epoch. Tail shapes compile once. Drop never submits or waits. Different models/configurations are rejected.

StRust public API

Declaration

pub struct PpoUpdateProgram { /* private fields */ }
Public Rust declaration.

Public Methods

pub fn invalidate_updates(&mut self)
pub fn replay_count(&self) -> u64
pub fn command_recording_count(&self) -> u64
pub fn is_compiled(&self) -> bool

Associated Functions

fn default() -> PpoUpdateProgram

Public Method Documentation

pub fn invalidate_updates( &mut self )
No public source comment is attached to this declaration.
pub fn replay_count( &self ) -> u64
No public source comment is attached to this declaration.
pub fn command_recording_count( &self ) -> u64
No public source comment is attached to this declaration.
pub fn is_compiled( &self ) -> bool
No public source comment is attached to this declaration.
fn default() -> PpoUpdateProgram
No public source comment is attached to this declaration.