keras-team/keras · error · ValueError

Incompatible shapes between `a` and `b`. Expected `a.shape[-

Error message

Incompatible shapes between `a` and `b`. Expected `a.shape[-1] == b.shape[-1]`. Received: a.shape={a.shape}, b.shape={b.shape}

What it means

In the rank(a) == rank(b) - 1 branch of _assert_a_b_compat, keras.ops.solve / keras.ops.solve_triangular treat b as a batch of vectors and require a.shape[-1] == b.shape[-1]: the number of unknowns must equal the length of each RHS vector. This error means the coefficient matrix's column count differs from the right-hand-side vector length.

Source

Thrown at keras/src/ops/linalg.py:875

        m, n = a.shape[-2:]
        if m != n:
            raise ValueError(
                "Expected a square matrix. "
                f"Received non-square input with shape {a.shape}"
            )


def _assert_a_b_compat(a, b):
    if a.ndim == b.ndim:
        if a.shape[-2] != b.shape[-2]:
            raise ValueError(
                "Incompatible shapes between `a` and `b`. "
                "Expected `a.shape[-2] == b.shape[-2]`. "
                f"Received: a.shape={a.shape}, b.shape={b.shape}"
            )
    elif a.ndim == b.ndim - 1:
        if a.shape[-1] != b.shape[-1]:
            raise ValueError(
                "Incompatible shapes between `a` and `b`. "
                "Expected `a.shape[-1] == b.shape[-1]`. "
                f"Received: a.shape={a.shape}, b.shape={b.shape}"
            )


class JVP(Operation):
    def __init__(self, has_aux=False, *, name=None):
        super().__init__(name=name)
        self.has_aux = has_aux

    def call(self, fun, primals, tangents):
        """Computes the JVP of `fun` at `primals` along `tangents`.

        Args:
            fun: A callable that takes tensors (or nested structures) as input
                 and returns a tensor (or nested structure) as output.
            primals: Input tensors (or nested structures) at which the Jacobian

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Align dimensions: build b with b.shape[-1] == a.shape[-1]; for multiple RHS use shape (n, k) so ranks match and the row rule applies.
2. Audit where b is produced and ensure it is not a slice/padding artifact with a different length than the system size.
3. Add a pre-call check: assert a.shape[-1] == b.shape[-1] (vector case) or a.shape[-2] == b.shape[-2] (matrix case).

Example fix

// before
from keras import ops
import numpy as np
A = np.random.rand(5, 5)
b = np.random.rand(7)     # length 7, but 5 unknowns
x = ops.solve(A, b)       # ValueError

// after
A = np.random.rand(5, 5)
b = np.random.rand(5)     # one entry per unknown
x = ops.solve(A, b)

from keras import ops

def check_solve_vector_case(a, b):
    if a.ndim == b.ndim - 1:
        sa, sb = ops.shape(a)[-1], ops.shape(b)[-1]
        assert sa is None or sb is None or sa == sb, (
            f"a.shape[-1]={sa} != b.shape[-1]={sb}")

check_solve_vector_case(A, b)
x = ops.solve(A, b)

def vector_rhs_matches(a, b) -> bool:
    return a.ndim == b.ndim - 1 and (
        a.shape[-1] is None or b.shape[-1] is None or a.shape[-1] == b.shape[-1]
    )

When it happens

Trigger: Calling keras.ops.solve(A, b) with A of shape (3, 3) and b of shape (4,) (rank differs by one), or batched A (B, 5, 5) with vectors b of shape (B, 4); using solve_triangular with a factor of size n but an RHS vector of length m != n.

Common situations: Solving square systems where the RHS was assembled from a different feature dimension; migrating from np.linalg.solve where NumPy raises its own mismatch error, making the Keras requirement non-obvious; feeding flattened labels of the wrong length as b.

Related errors

• Incompatible shapes between `a` and `b`. Expected `a.shape[-
• Expected input to have rank >= 2. Received input with shape
• Expected a square matrix. Received non-square input with sha
• The `weights` argument should be either `None` (random initi
• Expected mode to be one of `caffe`, `tf` or `torch`. Receive

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/c73ba75c09f27cbb. Report an issue: GitHub.