Struct AdamConfig 

pub struct AdamConfig {
    pub learning_rate: f32,
    pub beta1: f32,
    pub beta2: f32,
    pub eps: f32,
    pub weight_decay: f32,
    pub amsgrad: bool,
}

Configuration for the Adam optimization algorithm

Contains all hyperparameters that control the behavior of Adam optimization. Default values follow PyTorch conventions for maximum compatibility and optimal convergence across a wide range of neural network architectures.

Fields

• learning_rate - Step size for parameter updates (default: 1e-3)
• beta1 - Exponential decay rate for first moment estimates (default: 0.9)
• beta2 - Exponential decay rate for second moment estimates (default: 0.999)
• eps - Small constant for numerical stability (default: 1e-8)
• weight_decay - L2 regularization coefficient (default: 0.0)
• amsgrad - Whether to use AMSGrad variant for improved stability (default: false)

Fields

learning_rate: f32

Learning rate for parameter updates (default: 1e-3)

beta1: f32

Exponential decay rate for first moment estimates (default: 0.9)

beta2: f32

Exponential decay rate for second moment estimates (default: 0.999)

eps: f32

Small constant for numerical stability (default: 1e-8)

weight_decay: f32

Weight decay coefficient for L2 regularization (default: 0.0)

amsgrad: bool

Whether to use AMSGrad variant (default: false)

Trait Implementations

impl Clone for AdamConfig

fn clone(&self) -> AdamConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AdamConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for AdamConfig

fn default() -> Self

Returns the “default value” for a type.

impl FromFieldValue for AdamConfig

fn from_field_value( value: FieldValue, field_name: &str, ) -> SerializationResult<Self>

Create AdamConfig from FieldValue with comprehensive validation

This method reconstructs an AdamConfig instance from a FieldValue::Object, performing type validation and field extraction. It’s designed to handle AdamConfig instances that were embedded as fields within larger serializable structures, ensuring proper error handling and detailed error messages.

Arguments

• value - FieldValue containing configuration data (must be Object variant)
• field_name - Name of the field being deserialized for error context

Returns

Reconstructed AdamConfig instance on success, or SerializationError on failure

Expected FieldValue Structure

The FieldValue must be an Object variant containing:

• “learning_rate”: Numeric field value
• “beta1”: Numeric field value
• “beta2”: Numeric field value
• “eps”: Numeric field value
• “weight_decay”: Numeric field value
• “amsgrad”: Boolean field value

Errors

Returns SerializationError if:

• FieldValue is not an Object variant
• Any required field is missing from the object
• Any field has incorrect type or invalid value
• Deserialization process fails for any reason

Performance

• Time Complexity: O(1) - Constant time field extraction and validation
• Memory Usage: Temporary deserializer allocation for field processing
• Error Handling: Detailed error messages with field name context

impl StructSerializable for AdamConfig

fn to_serializer(&self) -> StructSerializer

Convert AdamConfig to StructSerializer for comprehensive serialization

This method serializes all Adam hyperparameters into a structured format suitable for both JSON and binary serialization. Every field is essential for proper optimizer reconstruction and training continuation. The serialization preserves exact floating-point values and boolean flags to ensure identical behavior after deserialization.

Returns

StructSerializer containing all configuration data with field names and values

Serialized Fields

• learning_rate: Base learning rate for parameter updates
• beta1: Exponential decay rate for first moment estimates
• beta2: Exponential decay rate for second moment estimates
• eps: Small constant for numerical stability in denominator
• weight_decay: L2 regularization coefficient
• amsgrad: Boolean flag for AMSGrad variant usage

Performance

• Time Complexity: O(1) - Constant time field serialization
• Memory Usage: Minimal allocation for field storage
• Precision: Full floating-point precision preservation

fn from_deserializer( deserializer: &mut StructDeserializer, ) -> SerializationResult<Self>

Create AdamConfig from StructDeserializer with full validation

This method reconstructs an AdamConfig instance from serialized hyperparameters, performing comprehensive validation to ensure all required fields are present and contain valid values. The deserialization process maintains exact floating-point precision and validates that all hyperparameters are within reasonable ranges.

Arguments

• deserializer - StructDeserializer containing configuration field data

Returns

Reconstructed AdamConfig instance on success, or SerializationError on failure

Required Fields

All fields must be present in the deserializer:

• learning_rate: Must be a valid f32 value
• beta1: Must be a valid f32 value (typically 0.0-1.0)
• beta2: Must be a valid f32 value (typically 0.0-1.0)
• eps: Must be a valid f32 value (typically small positive)
• weight_decay: Must be a valid f32 value (typically 0.0 or small positive)
• amsgrad: Must be a valid boolean value

Errors

Returns SerializationError if:

• Any required field is missing from the deserializer
• Any field contains invalid data type
• Field extraction fails for any reason

Performance

• Time Complexity: O(1) - Constant time field extraction
• Memory Usage: Minimal allocation for configuration structure
• Validation: Comprehensive field presence and type validation

fn save_json<P: AsRef<Path>>(&self, path: P) -> SerializationResult<()>

Saves the struct to a JSON file

fn save_binary<P: AsRef<Path>>(&self, path: P) -> SerializationResult<()>

Saves the struct to a binary file

fn load_json<P: AsRef<Path>>(path: P) -> SerializationResult<Self>

Loads the struct from a JSON file

fn load_binary<P: AsRef<Path>>(path: P) -> SerializationResult<Self>

Loads the struct from a binary file

fn to_json(&self) -> SerializationResult<String>

Converts the struct to a JSON string

fn to_binary(&self) -> SerializationResult<Vec<u8>>

Converts the struct to binary data

fn from_json(json: &str) -> SerializationResult<Self>

Creates the struct from a JSON string

fn from_binary(data: &[u8]) -> SerializationResult<Self>

Creates the struct from binary data

impl ToFieldValue for AdamConfig

fn to_field_value(&self) -> FieldValue

Convert AdamConfig to FieldValue for embedding in larger structures

This method converts the AdamConfig into a FieldValue::Object that can be embedded as a field within larger serializable structures. This enables AdamConfig to be serialized as part of more complex training configurations or model checkpoints while maintaining its structured representation.

Returns

FieldValue::Object containing all configuration data as key-value pairs

Object Structure

The returned object contains these fields:

• “learning_rate”: f32 value as FieldValue::F32
• “beta1”: f32 value as FieldValue::F32
• “beta2”: f32 value as FieldValue::F32
• “eps”: f32 value as FieldValue::F32
• “weight_decay”: f32 value as FieldValue::F32
• “amsgrad”: bool value as FieldValue::Bool

Performance

• Time Complexity: O(1) - Constant time field conversion
• Memory Usage: Allocates HashMap for field storage
• Conversion: Direct field-to-FieldValue conversion without copying