reduce

virtual void reduce(idx_t n, const float *x, float *res) const = 0;

API for dimension reduction. No specific implementation is provided.

idx_t n: number of inputs for inference.

const float *x: feature vector to be inferred.

float* res: feature vector result obtained after inference.

N/A

• Range of n: 0 < n < 1e9.
• x and res must be non-null pointers, and their lengths must be dimIn × n and dimOut × n, respectively. Otherwise, an out-of-bounds read/write error may occur, causing program breakdown.