tracel-ai/burn · error

NdArray supports arrays up to 6 dimensions, received: {}

Error message

NdArray supports arrays up to 6 dimensions, received: {}

What it means

Dimension-limit guard in burn-ndarray's reshape macro: ndarray-backed tensors can hold at most 6 dimensions; the macro's match expands ranks 1–6 and panics for any D beyond that. It fires when a tensor op on the ndarray backend is asked to produce/reshape an array of 7+ dimensions — the failing input is the over-rank shape.

Source

Thrown at crates/burn-ndarray/src/tensor.rs:582

            },
            false => $array.to_shape(dim).unwrap().as_standard_layout().into_shared(),
        };
        array.into_dyn()
    }};
    (
        ty $ty:ty,
        shape $shape:expr,
        array $array:expr,
        d $D:expr
    ) => {{
        match $D {
            1 => reshape!(ty $ty, n 1, shape $shape, array $array),
            2 => reshape!(ty $ty, n 2, shape $shape, array $array),
            3 => reshape!(ty $ty, n 3, shape $shape, array $array),
            4 => reshape!(ty $ty, n 4, shape $shape, array $array),
            5 => reshape!(ty $ty, n 5, shape $shape, array $array),
            6 => reshape!(ty $ty, n 6, shape $shape, array $array),
            _ => core::panic!("NdArray supports arrays up to 6 dimensions, received: {}", $D),
        }
    }};
}

/// Slice a tensor
#[macro_export]
macro_rules! slice {
    ($tensor:expr, $slices:expr) => {
        slice!($tensor, $slices, F64, F32, I64, I32, I16, I8, U64, U32, U16, U8, Bool)
    };
    ($tensor:expr, $slices:expr, $($variant:ident),*) => {
        match $tensor {
            $(NdArrayTensor::$variant(s) => { NdArrayOps::slice(s.view(), $slices).into() })*
        }
    };
}

impl NdArrayTensor {

View on GitHub (pinned to d16f7ba2ed)

Solutions

1. Restructure the model to keep tensors at rank <= 6 (merge axes, e.g. fold batch*time into one dimension)
2. Merge adjacent dimensions before the op and re-expand afterwards
3. Switch to a backend without this rank limit (e.g. burn-torch) if high-rank ops are required

Example fix

// before
let x = x.reshape([2, 2, 2, 2, 2, 2, 2]); // 7-D -> panic
// after
let x = x.reshape([2, 2, 2, 2, 2, 4]); // merged last two axes, rank 6

assert!(new_shape.len() <= 6, "ndarray backend supports max rank 6");

fn rank_supported(shape: &[usize]) -> bool {
    (1..=6).contains(&shape.len())
}

When it happens

Trigger: Reshaping a tensor to (or operating on a tensor of) rank > 6, e.g. reshape to 7+ dimensions via tensor.reshape([...7+ dims...]) on burn-ndarray.

Common situations: Models stacking many axes (video batch x time x frames x channels x ...), nested loops of unsqueeze, or generic code building shapes dynamically that can exceed 6.

Related errors

• Matrix multiplication requires an array with at least 2 dime
• Dimensions are incompatible for matrix multiplication: LHS c
• Shape should be compatible shape={dim:?}: {err:?}
• broadcast_shape: incompatible dimensions {} and {} at positi
• Dimensions differ and cannot be broadcasted.

AI-assisted analysis of tracel-ai/burn@d16f7ba2ed (2026-09-05). Data as JSON: /api/errors/53d096e9197eb941. Report an issue: GitHub.