Struct NfspIterationStats 

pub struct NfspIterationStats {
    pub iteration: usize,
    pub reservoir_sizes: Vec<usize>,
    pub br_stats: Option<JointStats>,
    pub avg_policy_loss: Vec<Option<f64>>,
    pub br_action_marginal: Vec<Option<Vec<f32>>>,
    pub avg_action_marginal: Vec<Option<Vec<f32>>>,
    pub cumulative_br_pushes: usize,
    pub cumulative_rollout_steps: usize,
}

Per-iteration NFSP statistics.

Fields

iteration: usize

Iteration index (1-based).

reservoir_sizes: Vec<usize>

Per-agent reservoir size at the end of the iteration.

br_stats: Option<JointStats>

Per-agent best-response training stats (the last joint-trainer call’s JointStats).

avg_policy_loss: Vec<Option<f64>>

Per-agent average-policy cross-entropy loss (averaged across AP supervised steps). None if the AP was skipped because the reservoir was empty.

br_action_marginal: Vec<Option<Vec<f32>>>

Per-agent BR-action marginal under the constant matching- pennies observation. None when the trainer is not on a constant-obs env (callers may ignore on non-matching-pennies envs).

avg_action_marginal: Vec<Option<Vec<f32>>>

Per-agent AP-action marginal under the constant matching- pennies observation. None when the trainer is not on a constant-obs env.

cumulative_br_pushes: usize

Cumulative count of BR-sampled actions (across the η-mixing rollout collector) since the trainer was constructed.

cumulative_rollout_steps: usize

Cumulative count of rollout steps observed (across all agents).

Trait Implementations

impl Clone for NfspIterationStats

fn clone(&self) -> NfspIterationStats

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for NfspIterationStats

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

Formats the value using the given formatter.

impl Default for NfspIterationStats

fn default() -> NfspIterationStats

Returns the “default value” for a type.