Struct TrainerConfig

pub struct TrainerConfig {
    pub max_opts: usize,
    pub opt_interval: usize,
    pub eval_interval: usize,
    pub flush_record_interval: usize,
    pub record_compute_cost_interval: usize,
    pub record_agent_info_interval: usize,
    pub warmup_period: usize,
    pub save_interval: usize,
}

Configuration parameters for the training process.

This struct defines various intervals and thresholds that control the behavior of the training loop. Each parameter can be set using the builder pattern methods.

Fields

max_opts: usize

Maximum number of optimization steps to perform. Training stops when this number is reached.

opt_interval: usize

Number of environment steps between optimization updates. For example, if set to 1, optimization occurs after every environment step.

eval_interval: usize

Number of optimization steps between performance evaluations. During evaluation, the agent’s performance is measured and the best model is saved.

flush_record_interval: usize

Number of optimization steps between flushing recorded metrics to storage. This controls how frequently training metrics are persisted.

record_compute_cost_interval: usize

Number of optimization steps between recording computational performance metrics. This includes metrics like optimization steps per second.

record_agent_info_interval: usize

Number of optimization steps between recording agent-specific information. This can include internal agent metrics or state information.

warmup_period: usize

Initial number of environment steps before optimization begins. During this period, the replay buffer is filled with initial experiences.

save_interval: usize

Number of optimization steps between saving model checkpoints. These checkpoints can be used for resuming training or analysis.

Implementations

impl TrainerConfig

pub fn max_opts(self, v: usize) -> Self

Sets the maximum number of optimization steps.

Arguments

• v - Maximum number of optimization steps

Returns

Self with the updated configuration

pub fn eval_interval(self, v: usize) -> Self

Sets the interval between performance evaluations.

Arguments

• v - Number of optimization steps between evaluations

Returns

Self with the updated configuration

pub fn eval_threshold(self, _v: f32) -> Self

(Deprecated) Sets the evaluation threshold.

This method is currently unimplemented and may be removed in future versions.

pub fn opt_interval(self, opt_interval: usize) -> Self

Sets the interval between optimization updates.

Arguments

• opt_interval - Number of environment steps between optimizations

Returns

Self with the updated configuration

pub fn flush_record_interval(self, flush_record_interval: usize) -> Self

Sets the interval for flushing recorded metrics.

Arguments

• flush_record_interval - Number of optimization steps between flushes

Returns

Self with the updated configuration

pub fn record_compute_cost_interval( self, record_compute_cost_interval: usize, ) -> Self

Sets the interval for recording computational performance metrics.

Arguments

• record_compute_cost_interval - Number of optimization steps between recordings

Returns

Self with the updated configuration

pub fn record_agent_info_interval( self, record_agent_info_interval: usize, ) -> Self

Sets the interval for recording agent-specific information.

Arguments

• record_agent_info_interval - Number of optimization steps between recordings

Returns

Self with the updated configuration

pub fn warmup_period(self, warmup_period: usize) -> Self

Sets the initial warmup period before optimization begins.

Arguments

• warmup_period - Number of environment steps in the warmup period

Returns

Self with the updated configuration

pub fn save_interval(self, save_interval: usize) -> Self

Sets the interval for saving model checkpoints.

Arguments

• save_interval - Number of optimization steps between checkpoints

Returns

Self with the updated configuration

pub fn load(path: impl AsRef<Path>) -> Result<Self>

Loads configuration from a YAML file.

Arguments

• path - Path to the configuration file

Returns

Result containing the loaded configuration

Errors

Returns an error if the file cannot be read or parsed

pub fn save(&self, path: impl AsRef<Path>) -> Result<()>

Saves the configuration to a file.

Arguments

• path - Path where the configuration will be saved

Returns

Result indicating success or failure

Errors

Returns an error if the file cannot be written

Trait Implementations

impl Clone for TrainerConfig

fn clone(&self) -> TrainerConfig

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Debug for TrainerConfig

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

Formats the value using the given formatter.

impl Default for TrainerConfig

fn default() -> Self

Creates a default configuration with conservative values.

Default values are set to:

• max_opts: 0
• opt_interval: 1 (optimize every step)
• eval_interval: 0 (no evaluation)
• flush_record_interval: usize::MAX (never flush)
• record_compute_cost_interval: usize::MAX (never record)
• record_agent_info_interval: usize::MAX (never record)
• warmup_period: 0 (no warmup)
• save_interval: usize::MAX (never save)

impl<'de> Deserialize<'de> for TrainerConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for TrainerConfig

fn eq(&self, other: &TrainerConfig) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for TrainerConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for TrainerConfig