LayerActivationStats<T>
Models & Types · AiDotNet.Deployment.Optimization.Quantization.Calibration
Activation statistics for a single layer.
Constructors
| Constructor |
Summary |
LayerActivationStats() |
|
Properties
| Property |
Summary |
double MaxAbsValue { get; set; } |
Maximum absolute activation value observed. |
double MaxValue { get; set; } |
Maximum activation value observed. |
double Mean { get; set; } |
Running mean of activation values. |
double MinValue { get; set; } |
Minimum activation value observed. |
double StandardDeviation { get; } |
Gets the standard deviation of activations. |
double Variance { get; set; } |
Running variance of activation values (for standard deviation). |
double[] PerChannelMaxAbs { get; set; } |
Per-channel maximum absolute values (for per-channel quantization). |
int SampleCount { get; set; } |
Number of samples accumulated. |
string LayerName { get; set; } |
Layer name or identifier. |
Methods
| Method |
Summary |
void Update(Tensor<T> activations) |
Updates statistics with a new batch of activations. |