jax-ml/jax · error · ValueError

expected deg >= 0

Error message

expected deg >= 0

What it means

jnp.polyfit(x, y, deg) performs a least-squares polynomial fit and needs deg to be a non-negative Python integer (it is converted via core.concrete_or_error(int, ...)). A negative degree has no meaning — there is no polynomial of degree -1 — so it raises ValueError before building the Vandermonde matrix.

Source

Thrown at jax/_src/numpy/polynomial.py:237

    s: [1.67 0.47 0.04]
    rcond: 4.7683716e-07

    If ``cov=True`` and ``full=False``, returns a tuple of arrays having
    polynomial coefficients and covariance matrix.

    >>> p, C = jnp.polyfit(x, y, 2, cov=True)
    >>> p.shape, C.shape
    ((3, 3), (3, 3, 3))
  """
  if w is None:
    x_arr, y_arr = ensure_arraylike("polyfit", x, y)
  else:
    x_arr, y_arr, w = ensure_arraylike("polyfit", x, y, w)
  del x, y
  deg = core.concrete_or_error(int, deg, "deg must be int")
  order = deg + 1
  if deg < 0:
    raise ValueError("expected deg >= 0")
  if x_arr.ndim != 1:
    raise TypeError("expected 1D vector for x")
  if x_arr.size == 0:
    raise TypeError("expected non-empty vector for x")
  if y_arr.ndim < 1 or y_arr.ndim > 2:
    raise TypeError("expected 1D or 2D array for y")
  if x_arr.shape[0] != y_arr.shape[0]:
    raise TypeError("expected x and y to have same length")

  if rcond is None:
    rcond = len(x_arr) * float(finfo(x_arr.dtype).eps)
  rcond = core.concrete_or_error(float, rcond, "rcond must be float")
  # set up least squares equation for powers of x
  lhs = vander(x_arr, order)
  rhs = y_arr

  # apply weighting
  if w is not None:

View on GitHub (pinned to 1e1c6a8fc0)

Solutions

1. Clamp/validate deg: use max(deg, 0) only if that's actually intended; otherwise fix the source of the negative value.
2. Check the order-vs-degree convention: order = deg + 1, so deg = n_coeffs - 1.
3. Pass a static Python int, not a traced value, when under jit.

Example fix

// before
coeffs = jnp.polyfit(x, y, deg=n_coeffs - 1)  # n_coeffs == 0
// after
coeffs = jnp.polyfit(x, y, deg=max(n_coeffs - 1, 0))
// or assert n_coeffs >= 1 before the call

deg = int(deg)
if deg < 0:
    raise ValueError(f'deg must be >= 0, got {deg}')
c = jnp.polyfit(x, y, deg)

def valid_deg(d) -> bool:
    return isinstance(d, int) and not isinstance(d, bool) and d >= 0

When it happens

Trigger: jnp.polyfit(x, y, deg=-1); deg computed as order-1 where order=0; deg passed as a tracer under jit (sibling concrete_or_error failure).

Common situations: Off-by-one when converting between 'number of coefficients' (order) and 'degree' (order-1); looping over degrees and including 0 or negatives; config-driven degree values validated only as ints, not as >= 0.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

• expected 1D vector for x
• expected non-empty vector for x
• expected 1D or 2D array for y
• expected x and y to have same length
• expected a 1-d array for weights

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/49cb720adfd2da5d. Report an issue: GitHub.