keras-team/keras · error · ValueError

`dot_product_attention` only supports 4D inputs. Received: q

Error message

`dot_product_attention` only supports 4D inputs. Received: query.shape={query.shape}, key.shape={key.shape}, value.shape={value.shape}.

What it means

Error "`dot_product_attention` only supports 4D inputs. Received: query.shape={query.shape}, key.shape={key.shape}, value.shape={value.shape}." thrown in keras-team/keras.

Source

Thrown at keras/src/backend/jax/nn.py:1687

        scale: Float. Optional scale that is applied to the attention
            computation.
        is_causal: Boolean. Specifying whether causal masking is applied.
        flash_attention: Boolean. Whether to use flash attention optimization
            for increased performance. Default to None, which means it will
            be auto-determined based on the platform, input shapes and
            compatibility.
        attn_logits_soft_cap: Float. Optional float to softly cap attention
            logits to avoid numerical stability issues. Applied as:
            `logits = logits / (1.0 + abs(logits) / attn_logits_soft_cap)`.

    Returns:
        JAX Array of shape `[batch, time, heads, depth_v]`.
    """
    query = convert_to_tensor(query)
    key = convert_to_tensor(key)
    value = convert_to_tensor(value)
    if len(query.shape) != 4 or len(key.shape) != 4 or len(value.shape) != 4:
        raise ValueError(
            "`dot_product_attention` only supports 4D inputs. "
            f"Received: query.shape={query.shape}, key.shape={key.shape}, "
            f"value.shape={value.shape}."
        )
    compute_dtype = backend.result_type(query.dtype, key.dtype, value.dtype)
    query = cast(query, compute_dtype)
    key = cast(key, compute_dtype)
    value = cast(value, compute_dtype)
    if bias is not None:
        bias = convert_to_tensor(bias, dtype=compute_dtype)

    # Check platform
    platform = jax.devices()[0].platform
    is_tpu = platform == "tpu"

    # Determine flash attention compatibility
    if flash_attention is None:
        flash_attention = _can_use_flash_attention(query, key, value, bias)

View on GitHub (pinned to 7a34a03db6)

When it happens

Trigger: Thrown at keras/src/backend/jax/nn.py:1687 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/289fd9b16824b0ec. Report an issue: GitHub.