keras-team/keras · error · ValueError

Cannot merge tensors with different batch sizes. Received te

Error message

Cannot merge tensors with different batch sizes. Received tensors with shapes {input_shape}

What it means

Merge layers require all input tensors to share the same batch dimension (None wildcard allowed). build() collects the first element of each input shape and raises if more than one distinct non-None batch size appears, because element-wise merging across different batch sizes is undefined.

Source

Thrown at keras/src/layers/merging/base_merge.py:117

    def build(self, input_shape):
        # Used purely for shape validation.
        if not isinstance(input_shape[0], (tuple, list)):
            raise ValueError(
                "A merge layer should be called on a list of inputs. "
                f"Received: input_shape={input_shape} (not a list of shapes)"
            )
        if len(input_shape) < 1:
            raise ValueError(
                "A merge layer should be called "
                "on a list of at least 1 input. "
                f"Received {len(input_shape)} inputs. "
                f"Full input_shape received: {input_shape}"
            )

        batch_sizes = {s[0] for s in input_shape if s} - {None}
        if len(batch_sizes) > 1:
            raise ValueError(
                "Cannot merge tensors with different batch sizes. "
                f"Received tensors with shapes {input_shape}"
            )

        if input_shape[0] is None:
            output_shape = None
        else:
            output_shape = input_shape[0][1:]

        for i in range(1, len(input_shape)):
            if input_shape[i] is None:
                shape = None
            else:
                shape = input_shape[i][1:]
            output_shape = self._compute_elemwise_op_output_shape(
                output_shape, shape
            )

View on GitHub (pinned to 7a34a03db6)

Solutions

Make batch dimensions consistent: use None (dynamic) batch in all Input definitions
Slice or pad one tensor so both have the same number of samples per batch
Remove batch_size=... hardcoding on Inputs feeding the merge

Example fix

# before
a = keras.Input(shape=(16,), batch_size=32)
b = keras.Input(shape=(16,))  # runtime batch 64
out = layers.Add()([a, b])  # ValueError

# after
a = keras.Input(shape=(16,))
out = layers.Add()([a, b])

Defensive patterns

Strategy: validation

Validate before calling

batch_sizes = {tuple(t.shape)[0] for t in inputs if len(t.shape)} - {None}
assert len(batch_sizes) <= 1, f'conflicting batch sizes: {batch_sizes}'

Type guard

def same_batch_size(shapes) -> bool:
    bs = {s[0] for s in shapes if s} - {None}
    return len(bs) <= 1

Prevention

Prefer dynamic (None) batch dimensions in Input definitions
Verify dataset batch sizes across modalities before merging
Avoid hardcoding batch_size when downstream merges exist

When it happens

Trigger: Calling Add()([x, y]) where x has batch size 32 and y has batch size 64; merging a fixed-batch Input with a dynamic-batch Input is allowed, but 32 vs 64 fails.

Common situations: Hardcoded batch dimensions in Input(shape=..., batch_size=32) mixing with dynamic batches; slicing one branch to a different number of samples; data pipelines producing mismatched batch sizes across modalities.

Related errors

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/b9ac6590d51deefd. Report an issue: GitHub.