ParameterRegistry<T>
Models & Types · AiDotNet.ModelLoading
Manages named parameters for weight loading.
For Beginners
This class is like a phone book for model parameters. Each parameter has a name (like “encoder.conv1.weight”) and we can look up or set parameters by their names.
When loading pretrained weights, we need to know:
- What parameters exist in our model
- What shape each parameter should be
- Where to actually put the weight data
This registry provides all three capabilities.
Constructors
| Constructor | Summary |
|---|---|
ParameterRegistry() |
Initializes a new empty parameter registry. |
Properties
| Property | Summary |
|---|---|
int Count { get; } |
Gets the number of registered parameters. |
Methods
| Method | Summary |
|---|---|
IEnumerable<string> GetNames() |
Gets all registered parameter names. |
WeightLoadResult Load(Dictionary<string, Tensor<T>> weights, Func<string, string> mapping, bool strict) |
Loads weights from a dictionary. |
WeightLoadValidation Validate(IEnumerable<string> weightNames, Func<string, string> mapping) |
Validates weights against registered parameters. |
bool TryGet(string name, Tensor<T> tensor) |
Tries to get a parameter by name. |
bool TrySet(string name, Tensor<T> value) |
Sets a parameter by name. |
int[] GetShape(string name) |
Gets the expected shape for a parameter. |
void Register(string name, int[] shape, Func<Tensor<T>> getter, Action<Tensor<T>> setter) |
Registers a parameter with getter and setter delegates. |
void RegisterChild(string prefix, ParameterRegistry<T> child) |
Registers a child ParameterRegistry with a prefix. |
void RegisterLayer(string prefix, ILayer<T> layer) |
Registers a layer’s weights and biases. |
void RegisterLayers(string prefix, IEnumerable<ValueTuple<string, ILayer<T>>> layers) |
Registers multiple layers with a naming pattern. |