WeightAnalysisResult<T>

Models & Types · AiDotNet.ModelCompression

Analysis results for model weights to guide compression decisions.

Constructors

Constructor Summary
WeightAnalysisResult()

Properties

Property Summary
CompressionType RecommendedTechnique { get; set; } Gets or sets the recommended compression technique based on analysis.
Dictionary<string, object> RecommendedParameters { get; set; } Gets or sets the recommended hyperparameters for the compression technique.
T ClusteringPotential { get; set; } Gets or sets the clustering potential (how much quantization will help).
T Entropy { get; set; } Gets or sets the entropy of weight distribution (for Huffman coding potential).
T EstimatedCompressionRatio { get; set; } Gets or sets the estimated compression ratio achievable.
T MaxMagnitude { get; set; } Gets or sets the maximum weight magnitude.
T MeanMagnitude { get; set; } Gets or sets the mean weight magnitude.
T MinMagnitude { get; set; } Gets or sets the minimum weight magnitude.
T PruningPotential { get; set; } Gets or sets the fraction of weights that are near-zero (pruning potential).
T StdDevMagnitude { get; set; } Gets or sets the standard deviation of weight magnitudes.
long NearZeroWeights { get; set; } Gets or sets the number of weights with near-zero magnitude.
long TotalWeights { get; set; } Gets or sets the total number of weights analyzed.
long UniqueValues { get; set; } Gets or sets the number of unique weight values (for clustering potential).
string RecommendationReasoning { get; set; } Gets or sets the reasoning for the recommendation.