pub fn cuda_gather_dispatch(
data: &DeviceTensor,
indices: &DeviceTensor,
axis: usize,
output: &DeviceTensor,
) -> TractResult<()>pub fn cuda_gather_dispatch(
data: &DeviceTensor,
indices: &DeviceTensor,
axis: usize,
output: &DeviceTensor,
) -> TractResult<()>