Skip to main content

Module gradient_checkpointer

Module gradient_checkpointer 

Source
Expand description

GradientCheckpointer — gradient accumulation, checkpointing, and replay for distributed training with fault tolerance.

This module implements a production-grade gradient checkpointing system that supports configurable accumulation modes (Sum, Mean, WeightedMean), L2-norm gradient clipping, FNV-1a content checksums, and bounded checkpoint history with automatic eviction.

Structs§

CheckpointId
Newtype wrapper for checkpoint identifiers.
CheckpointerConfig
Configuration for a GradientCheckpointer.
GcCheckpointerStats
Snapshot of checkpointer statistics.
GcGradientCheckpoint
A snapshot of accumulated and (optionally clipped) gradients at a particular training step.
GcGradientTensor
A gradient tensor associated with a named layer at a particular training step.
GradientCheckpointer
Production-grade gradient checkpointing engine.

Enums§

GcAccumulationMode
Controls how multiple gradient tensors for the same layer are combined during a flush.
GradientCheckpointerError
Errors produced by GradientCheckpointer.

Functions§

fnv1a_f64_slice
Compute FNV-1a hash over a slice of f64 values.