Expand description
Memory-efficient operations module Memory-efficient implementations for neural networks
This module provides memory optimization techniques including:
- Gradient checkpointing for reduced memory usage during training
- In-place operations to minimize memory allocations
- Memory pool management for efficient tensor allocation
- Memory-aware batch processing
- Lazy evaluation and computation graphs
Structsยง
- Batch
Processor Stats - Statistics for the batch processor
- Gradient
Checkpointing - Gradient checkpointing implementation for memory-efficient training
- InPlace
Operations - In-place operations manager for minimizing memory allocations
- Memory
Aware Batch Processor - Memory-aware batch processor for handling large datasets
- Memory
Efficient Layer - Memory-efficient neural network layer that processes data in chunks
- Memory
Pool - Memory pool for efficient tensor allocation and reuse
- Memory
Usage - Memory usage tracking and reporting
- Pool
Statistics - Statistics about the memory pool