xai-org/x-algorithm · error · ValueError

Unknown kernel

Error message

Unknown kernel: {kernel!r}

What it means

_cyclic_kernel_weights selects the smoothing kernel for cyclic time-phase embeddings: 'box' (default path returning _box_kernel), 'triangle', and 'cosine' are implemented. Any other kernel string falls through to ValueError, so misspelled or unsupported kernel names fail fast before embedding computation.

Solutions

1. Use 'box', 'triangle', or 'cosine'.
2. Fix typos/casing in the kernel config string.
3. Add a new kernel function and an if-branch if a different shape is required.

Example fix

# before
kernel: triangular

# after
kernel: triangle

assert kernel in {"box", "triangle", "cosine"}, kernel

def is_supported_kernel(k: str) -> bool:
    return k in {"box", "triangle", "cosine"}

When it happens

Trigger: Calling _compute_cyclic_kernel_embedding with kernel="gaussian", "tri", or "COSINE" (case-sensitive) via the feature-prep config.

Common situations: Feature configs ported from pipelines with more kernels; typos and casing mistakes in YAML.

Understand the failure class

Background: "Invalid value" and "allowed values are" config errors: what your library rejected and how to fix it — this error's family across 41 libraries.

Related errors

• attn_logit_cap_method
• Invalid attention implementation
• Rope type is not supported
• Unknown loss_type
• unknown scaling role

AI-assisted analysis of xai-org/x-algorithm@24c60942c5 (2026-08-28). Data as JSON: /api/errors/7a8f7e3c03cddc61. Report an issue: GitHub.

Appendix: source

Thrown at phoenix/xrex/models/recsys_feature_prep.py:98

def _box_kernel(u: jax.Array) -> jax.Array:
    return ((u >= -0.5) & (u < 0.5)).astype(jnp.float32)


def _triangle_kernel(u: jax.Array) -> jax.Array:
    return jnp.maximum(1.0 - jnp.abs(u), 0.0)


def _cyclic_kernel_weights(u: jax.Array, kernel: str, width: float) -> jax.Array:
    if kernel == "box":
        return _box_kernel(u)
    half_width = width / 2.0
    if kernel == "triangle":
        return _triangle_kernel(u / half_width)
    if kernel == "cosine":
        return _cosine_compact_kernel(u / half_width)
    raise ValueError(f"Unknown kernel: {kernel!r}")


def _compute_cyclic_time_phase(
    impr_ts_sec: jax.Array,
    period_seconds: int,
) -> tuple[jax.Array, jax.Array]:
    sec_in_period = jnp.mod(impr_ts_sec.astype(jnp.int32), period_seconds).astype(jnp.float32)
    phase = sec_in_period / float(period_seconds)
    valid_mask = impr_ts_sec > 0
    phase = jnp.where(valid_mask, phase, 0.0)
    return phase, valid_mask


def _compute_cyclic_kernel_embedding(
    phase: jax.Array,
    valid_mask: jax.Array,
    embedding_table: jax.Array,
    fprop_dtype: jnp.dtype,

View on GitHub (pinned to 24c60942c5)