tracel-ai/burn · error

prod_dim: unsupported dtype {:?}

Error message

prod_dim: unsupported dtype {:?}

What it means

burn-flex's `prod_dim` reduces a product along a dimension, supporting float dtypes and integer dtypes I8–I64 / U8–U64 with widening accumulators (identity 1). Unsupported dtypes like Bool or quantized hit the catch-all panic.

Source

Thrown at crates/burn-flex/src/ops/reduce.rs:420

            f16::from_f32,
        ),
        DType::BF16 => reduce_dim_half(
            &tensor,
            dim,
            1.0,
            |acc, x| acc * x,
            bf16::to_f32,
            bf16::from_f32,
        ),
        DType::I8 => reduce_dim_widening::<i8, _>(&tensor, dim, 1, |acc, x| acc.wrapping_mul(x)),
        DType::I16 => reduce_dim_widening::<i16, _>(&tensor, dim, 1, |acc, x| acc.wrapping_mul(x)),
        DType::I32 => reduce_dim_widening::<i32, _>(&tensor, dim, 1, |acc, x| acc.wrapping_mul(x)),
        DType::I64 => reduce_dim_impl::<i64, _>(&tensor, dim, 1, |acc, x| acc * x),
        DType::U8 => reduce_dim_widening::<u8, _>(&tensor, dim, 1, |acc, x| acc.wrapping_mul(x)),
        DType::U16 => reduce_dim_widening::<u16, _>(&tensor, dim, 1, |acc, x| acc.wrapping_mul(x)),
        DType::U32 => reduce_dim_widening::<u32, _>(&tensor, dim, 1, |acc, x| acc.wrapping_mul(x)),
        DType::U64 => reduce_dim_impl::<u64, _>(&tensor, dim, 1, |acc, x| acc * x),
        _ => panic!("prod_dim: unsupported dtype {:?}", tensor.dtype()),
    }
}

// ============================================================================
// Max / Min (all elements)
// ============================================================================

/// Max of all elements, returning a scalar tensor of shape \[1\].
pub fn max(tensor: FlexTensor) -> FlexTensor {
    // Asserted here rather than per dtype: every path seeds the fold with an infinity, so without
    // this they report that seed as the max of nothing instead of failing.
    assert!(
        tensor.layout().shape().num_elements() > 0,
        "max: cannot reduce an empty tensor"
    );
    match tensor.dtype() {
        DType::F32 => max_f32_reduce(&tensor),
        DType::F64 => float_extremum_f64_reduce::<true>(&tensor),

View on GitHub (pinned to d16f7ba2ed)

Solutions

1. Cast first: `mask.cast(DType::I32).prod_dim(1)`.
2. Use explicit boolean reduction semantics where possible after casting.
3. Dequantize quantized tensors before reducing.

Example fix

// before
let row_and = mask.prod_dim(1); // Bool
// after
let row_and = mask.cast(DType::I32).prod_dim(1);

assert!(!matches!(t.dtype(), DType::Bool | DType::QFloat(_)), "prod_dim unsupported for {:?}; cast or dequantize first", t.dtype());

fn is_prod_capable(d: DType) -> bool {
    matches!(d, DType::F32 | DType::F64 | DType::F16 | DType::BF16
        | DType::I8 | DType::I16 | DType::I32 | DType::I64
        | DType::U8 | DType::U16 | DType::U32 | DType::U64)
}

When it happens

Trigger: Calling `Tensor::prod_dim(dim)` on a Bool or quantized tensor — e.g. per-row logical AND of a mask implemented via product.

Common situations: Per-sequence conjunction over boolean masks without casting; product over quantized values; generic trait code receiving unexpected dtypes.

Related errors

• prod: unsupported dtype {:?}
• sum: unsupported dtype {:?}
• sum_dim: unsupported dtype {:?}
• mean_dim: unsupported dtype {:?}
• max: unsupported dtype {:?}

AI-assisted analysis of tracel-ai/burn@d16f7ba2ed (2026-09-05). Data as JSON: /api/errors/d1b1b8b943d2664a. Report an issue: GitHub.