tracel-ai/burn · error

reduce_bool_dim: unsupported dtype {:?}

Error message

reduce_bool_dim: unsupported dtype {:?}

What it means

Dtype-dispatch exhaustiveness panic in `reduce_bool_dim`, the shared float any/all-dim reduction: only float dtypes are handled; a non-float tensor reaching it panics. Since it is called by any_float_dim/all_float_dim, hitting this implies an op-dispatch bug rather than user error.

Source

Thrown at crates/burn-flex/src/ops/comparison.rs:1040

        DType::F64 => {
            let data: &[f64] = tensor.storage();
            reduce_bool_dim_with(&tensor, dim, init, combine, out_dtype, |idx| {
                data[idx] != 0.0
            })
        }
        DType::F16 => {
            let data: &[f16] = tensor.storage();
            reduce_bool_dim_with(&tensor, dim, init, combine, out_dtype, |idx| {
                data[idx].to_f32() != 0.0
            })
        }
        DType::BF16 => {
            let data: &[bf16] = tensor.storage();
            reduce_bool_dim_with(&tensor, dim, init, combine, out_dtype, |idx| {
                data[idx].to_f32() != 0.0
            })
        }
        _ => panic!("reduce_bool_dim: unsupported dtype {:?}", tensor.dtype()),
    }
}

/// Reduce along a dimension producing a bool tensor (for int any/all_dim).
fn reduce_bool_dim_int(
    tensor: &FlexTensor,
    dim: usize,
    init: bool,
    combine: fn(bool, bool) -> bool,
    out_dtype: BoolDType,
) -> FlexTensor {
    let tensor = tensor.to_contiguous();
    macro_rules! dispatch {
        ($ty:ty) => {{
            let data: &[$ty] = tensor.storage();
            reduce_bool_dim_with(&tensor, dim, init, combine, out_dtype, |idx| data[idx] != 0)
        }};
    }

View on GitHub (pinned to d16f7ba2ed)

Solutions

1. Use any_int_dim/all_int_dim for integer tensors
2. Cast the tensor to F32 before the dim-wise float reduction
3. Assert the dtype is a float variant before calling

Example fix

// before
let row_any = any_float_dim(u8_tensor, 1, BoolDType::Native);
// after
let row_any = any_int_dim(u8_tensor, 1, BoolDType::Native);

if !matches!(t.dtype(), DType::F32|DType::F64|DType::F16|DType::BF16) { /* use any_int_dim/all_int_dim or cast */ }

fn is_float_dtype(d: &DType) -> bool {
    matches!(d, DType::F32|DType::F64|DType::F16|DType::BF16)
}

When it happens

Trigger: Calling any_float_dim or all_float_dim with a tensor whose dtype is not F32/F64/F16/BF16 — typically an integer or bool tensor.

Common situations: Dim-wise any/all on masks stored as U8 or on int tensors; generic code that picks any_float_dim without inspecting dtype.

Related errors

• any_float: unsupported dtype {:?}
• all_float: unsupported dtype {:?}
• any_int: unsupported dtype {:?}
• all_int: unsupported dtype {:?}
• reduce_bool_dim_int: unsupported dtype {:?}

AI-assisted analysis of tracel-ai/burn@d16f7ba2ed (2026-09-05). Data as JSON: /api/errors/0979b1412cbe96c7. Report an issue: GitHub.