Function softmax_f32 

pub fn softmax_f32(
    input: &[f32],
    num_rows: usize,
    row_size: usize,
    output: &mut [f32],
) -> Result<()>

Softmax for 32-bit floating point tensors.