Function cross_entropy_loss_forward 

pub fn cross_entropy_loss_forward<T: Triton, const BLOCK_SIZE: i32>(
    input_ptr: T::Pointer<f32>,
    targets_ptr: T::Pointer<i32>,
    out_ptr: T::Pointer<f32>,
    _n_rows: i32,
    n_cols: i32,
)
where
    T::I32Tensor: Tensor<i32, 1> + Comparison<i32, BoolTensor = T::BoolTensor>,
    T::Pointer<f32>: AddOffsets<i32, 1, T::I32Tensor, Output = T::Tensor<T::Pointer<f32>>> + AddOffsets<i32, 1, T::Tensor<i32>, Output = T::Tensor<T::Pointer<f32>>>,
    T::Pointer<i32>: AddOffsets<i32, 1, T::I32Tensor, Output = T::Tensor<T::Pointer<i32>>>,
    T::Tensor<i32>: Tensor<i32, 1>,

Cross-entropy loss forward: out[n] = log(sum_c exp(x[n,c])) - x[n, target[n]].

Numerically stable: subtracts row-max before exp (log-sum-exp trick).

Grid: [n_rows, 1, 1] — one CTA per row. BLOCK_SIZE must equal next_power_of_two(n_cols).