pub fn bmm_f32(
lhs: &[f32],
rhs: &[f32],
batch: usize,
rows: usize,
columns: usize,
reduction: usize,
lhs_batch_stride: usize,
lhs_row_stride: usize,
rhs_batch_stride: usize,
rhs_row_stride: usize,
transpose_lhs: bool,
transpose_rhs: bool,
) -> Vec<f32>