Expand description
GPU acceleration and optimization features for embedding models
This module provides advanced GPU acceleration capabilities including memory pooling, tensor caching, mixed precision, and compute optimization with full SciRS2 integration for maximum performance.
Structs§
- Accelerator
Stats - Statistics for GPU accelerator
- Allocation
Stats - Memory allocation statistics
- Batch
Size Optimizer - Batch size optimizer for maximizing GPU utilization
- Batch
Size Optimizer Stats - Statistics from batch size optimization
- Cache
Stats - Cache statistics
- Defragmentation
Result - Results of memory defragmentation operation
- Dynamic
Shape Handler - Dynamic shape handling for variable-size inputs
- GpuAcceleration
Config - GPU acceleration configuration
- GpuAcceleration
Manager - Main GPU acceleration manager
- GpuMemory
Pool - GPU memory pool for efficient memory management
- GpuPerformance
Stats - GPU performance statistics
- Memory
Defragmenter - Memory defragmentation utilities
- Mixed
Precision Processor - Mixed precision training and inference
- Multi
Stream Processor - Multi-stream processor for parallel GPU operations
- OutOf
Core Processor - Out-of-core processing for handling datasets larger than GPU memory
- SciR
S2Gpu Accelerator - Advanced GPU accelerator using SciRS2’s full GPU capabilities
- Tensor
Cache - Tensor cache for frequently used tensors