Model Compression

28 public types in this namespace, organized by kind.

Models & Types (25)

Type Summary
CompressionAnalyzer<T> Analyzes model weights to determine optimal compression strategies.
CompressionFrontierPoint<T> One operating point on the compression-versus-accuracy frontier: the configured strategy applied to a magnitude-ranked fraction of the model’s weights, with the resulting model re-evaluated on real data.
CompressionMetrics<T> Provides metrics and statistics for model compression operations.
CompressionResult<T> Result of a compression operation.
DeepCompressionMetadata<T> Metadata for Deep Compression containing information from all three stages.
DeepCompressionStats Statistics about Deep Compression performance.
DeepCompression<T> Implements the Deep Compression algorithm from Han et al.
HuffmanEncodingCompression<T> Implements Huffman encoding compression for model weights using variable-length encoding.
HuffmanEncodingMetadata<T> Metadata for Huffman encoding compression.
HybridCompressionMetadata Legacy non-generic metadata for backward compatibility.
HybridCompressionMetadata<T> Metadata for hybrid compression combining clustering and Huffman encoding.
HybridHuffmanClusteringCompression<T> Implements hybrid compression that combines weight clustering with Huffman encoding.
LowRankFactorizationCompression<T> Implements Low-Rank Factorization compression using SVD-like decomposition.
LowRankFactorizationMetadata<T> Metadata for Low-Rank Factorization compression.
MatrixCompressionMetadata<T> Metadata for matrix compression operations that wraps the underlying vector compression metadata.
ModelCompressionReport<T> The model-compression audit produced by a configured compression strategy: the true size-versus-accuracy trade-off of the actual decompressed model, a sensitivity-aware Pareto frontier, and weight reconstruction error — everything needed to…
ProductQuantizationCompression<T> Implements Product Quantization (PQ) compression for model weights.
ProductQuantizationMetadata<T> Metadata for Product Quantization compression.
SparseCompressionResult<T> Result of sparse compression operation.
SparsePruningCompression<T> Implements sparse pruning compression by zeroing out small-magnitude weights.
SparsePruningMetadata<T> Metadata for sparse pruning compression.
TensorCompressionMetadata<T> Metadata for N-dimensional tensor compression operations that wraps the underlying vector compression metadata.
WeightAnalysisResult<T> Analysis results for model weights to guide compression decisions.
WeightClusteringCompression<T> Implements weight clustering compression using K-means clustering to group similar weights.
WeightClusteringMetadata<T> Metadata for weight clustering compression.

Base Classes (1)

Type Summary
ModelCompressionBase<T> Provides a base implementation for model compression techniques used to reduce model size while preserving accuracy.

Enums (1)

Type Summary
SparseFormat Sparse storage formats.

Helpers & Utilities (1)

Type Summary
HuffmanNode<T> Represents a node in the Huffman tree.