Struct SeqLenDistribution 

pub(super) struct SeqLenDistribution {
    pub min: usize,
    pub max: usize,
    pub mean: usize,
    pub p95: usize,
}

Per-batch token-length distribution statistics.

All lengths are measured in tokens (post-tokenization id counts), before any padding. Carried in super::types::EmbedStats and logged on every completed embed request so operators can correlate latency spikes with new (batch_size, seq_len) shapes hitting TRT engine compile paths.

Fields

min: usize

Minimum token sequence length in the batch.

max: usize

Maximum token sequence length in the batch.

mean: usize

Mean token sequence length (integer, truncated toward zero).

p95: usize

95th-percentile token sequence length.

Index is (n * 95) / 100 on a sorted copy — a conservative floor that never exceeds n - 1. For a batch of 64 texts this maps to index 60, meaning the 61st-shortest sequence.

Trait Implementations

impl Clone for SeqLenDistribution

fn clone(&self) -> SeqLenDistribution

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SeqLenDistribution

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

Formats the value using the given formatter.

impl Default for SeqLenDistribution

fn default() -> SeqLenDistribution

Returns the “default value” for a type.

impl Copy for SeqLenDistribution