keras-team/keras · error · ValueError

The shape of `window` must be equal to [sequence_length].Rec

Error message

The shape of `window` must be equal to [sequence_length].Received: window shape={win.shape}

What it means

Error "The shape of `window` must be equal to [sequence_length].Received: window shape={win.shape}" thrown in keras-team/keras.

Source

Thrown at keras/src/backend/jax/math.py:201

    x = convert_to_tensor(x)

    if center:
        pad_width = [(0, 0) for _ in range(len(x.shape))]
        pad_width[-1] = (fft_length // 2, fft_length // 2)
        x = jnp.pad(x, pad_width, mode="reflect")

    l_pad = (fft_length - sequence_length) // 2
    r_pad = fft_length - sequence_length - l_pad

    if window is not None:
        if isinstance(window, str):
            win = convert_to_tensor(
                scipy.signal.get_window(window, sequence_length), dtype=x.dtype
            )
        else:
            win = convert_to_tensor(window, dtype=x.dtype)
        if len(win.shape) != 1 or win.shape[-1] != sequence_length:
            raise ValueError(
                "The shape of `window` must be equal to [sequence_length]."
                f"Received: window shape={win.shape}"
            )
        win = jnp.pad(win, [[l_pad, r_pad]])
    else:
        win = jnp.ones((sequence_length + l_pad + r_pad), dtype=x.dtype)

    result = jax.scipy.signal.stft(
        x,
        fs=1.0,
        window=win,
        nperseg=(sequence_length + l_pad + r_pad),
        noverlap=(sequence_length + l_pad + r_pad - sequence_stride),
        nfft=fft_length,
        boundary=None,
        padded=False,
    )[-1]
    # scale and swap to (..., num_sequences, fft_bins)

View on GitHub (pinned to 7a34a03db6)

When it happens

Trigger: Thrown at keras/src/backend/jax/math.py:201 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/6779573ecf154437. Report an issue: GitHub.