pub fn group_gemm_f32(
lhs: &[f32],
rhs: &[f32],
rows: &[i32],
columns: &[i32],
reductions: &[i32],
lhs_offsets: &[i32],
rhs_offsets: &[i32],
output_offsets: &[i32],
output_len: usize,
transpose_rhs: bool,
) -> Vec<f32>