Method train

nodedb-vector/src/rerank/codecs/pq.rs:190–220

Train PQ codebooks via k-means on a sample of vectors. Validates that: - `samples` is non-empty. - Every sample has length `self.dim`. - `self.dim % self.m == 0` (PQ requires divisible dimensionality). - At least `self.k` samples are provided (k-means needs ≥ k points). On success, stores the trained codec and subsequent `encode` / `distance_prepared` calls will succeed.

(&mut self, samples: &[&[f32]])

    /// On success, stores the trained codec and subsequent `encode` /
    /// `distance_prepared` calls will succeed.
    fn train(&mut self, samples: &[&[f32]]) -> Result<(), RerankError> {
        if samples.is_empty() {
            return Err(RerankError::BadInput(
                "pq train: empty sample set".to_string(),
            ));
        }
        for s in samples {
            if s.len() != self.dim {
                return Err(RerankError::BadInput(format!(
                    "pq train: sample has len {} but codec dim is {}",
                    s.len(),
                    self.dim
                )));
            }
        }
        if !self.dim.is_multiple_of(self.m) {
            return Err(RerankError::BadInput(format!(
                "pq train: dim ({}) must be divisible by m ({})",
                self.dim, self.m
            )));
        }
        if samples.len() < self.k {
            return Err(RerankError::BadInput(format!(
                "pq train: need >= k samples for k-means, got {}",
                samples.len()
            )));
        }
        let codec = PqCodec::train(samples, self.dim, self.m, self.k, self.max_iter);
        self.codec = Some(codec);
        Ok(())
    }
}

// ── Tests ─────────────────────────────────────────────────────────────────────