jax-ml/jax · error · ValueError

Number of samples, {num}, must be non-negative.

Error message

Number of samples, {num}, must be non-negative.

What it means

The differentiable implementation behind jnp.linspace validates num (number of samples) and rejects negative values, since a negative sample count has no mathematical meaning and cannot be traced into a valid computation.

Source

Thrown at jax/_src/numpy/array_creation.py:596

    Array([[ 0.  ,  5.  ],
           [ 1.25,  6.25],
           [ 2.5 ,  7.5 ],
           [ 3.75,  8.75],
           [ 5.  , 10.  ]], dtype=float32)
  """
  num = core.concrete_dim_or_error(num, "'num' argument of jnp.linspace")
  axis = core.concrete_or_error(operator.index, axis, "'axis' argument of jnp.linspace")
  return _linspace(start, stop, num, endpoint, retstep, dtype, axis, device=device)

@api.jit(static_argnames=('num', 'endpoint', 'retstep', 'dtype', 'axis', 'device'))
def _linspace(start: ArrayLike, stop: ArrayLike, num: int = 50,
              endpoint: bool = True, retstep: bool = False,
              dtype: DTypeLike | None = None,
              axis: int = 0,
              *, device: xc.Device | Sharding | None = None) -> Array | tuple[Array, Array]:
  """Implementation of linspace differentiable in start and stop args."""
  if num < 0:
    raise ValueError(f"Number of samples, {num}, must be non-negative.")
  start, stop = util.ensure_arraylike("linspace", start, stop)

  if dtype is None:
    dtype = dtypes.to_inexact_dtype(dtypes.result_type(start, stop))
  else:
    dtype = dtypes.check_and_canonicalize_user_dtype(dtype, "linspace")
  computation_dtype = dtypes.to_inexact_dtype(dtype)
  start = start.astype(computation_dtype)
  stop = stop.astype(computation_dtype)

  bounds_shape = list(lax.broadcast_shapes(np.shape(start), np.shape(stop)))
  broadcast_start = util._broadcast_to(start, bounds_shape)
  broadcast_stop = util._broadcast_to(stop, bounds_shape)
  axis = len(bounds_shape) + axis + 1 if axis < 0 else axis
  bounds_shape.insert(axis, 1)
  div = (num - 1) if endpoint else num
  if num > 1:
    delta: Array = lax.convert_element_type(stop - start, computation_dtype) / asarray(div, dtype=computation_dtype)

View on GitHub (pinned to 1e1c6a8fc0)

Solutions

1. Clamp num: max(num, 0) or validate before calling
2. Recheck the count formula — linspace wants a count, not a step; use jnp.arange for step-based spacing

Example fix

# before
a = jnp.linspace(0.0, 1.0, num=(stop - start) // step)  # can be negative
# after
n = max(int((stop - start) // step), 0)
a = jnp.linspace(0.0, 1.0, num=n)

num = max(int(num), 0)
jnp.linspace(start, stop, num=num)

null

null

When it happens

Trigger: jnp.linspace(0, 1, num=-5) or num computed from an expression that can go negative (e.g. (b - a) // step).

Common situations: Computing sample counts from user parameters or deltas where rounding/division can produce -1 or lower; passing None-ish defaults that coerce to negative numbers.

Related errors

• `compute_on`'s compute_type argument must be a string.
• Argument '{x}' of type '{typ}' is not a valid JAX type
• Argument '{arg}' of type {type(arg)} is not a valid JAX type
• Expected kind to be a dtype, string, or tuple; got {kind=}
• at least one array or dtype is required

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/195ff19ee39ad0cb. Report an issue: GitHub.