Revert irrelevant changes

faster_whisper/feature_extractor.py

@@ -228,65 +228,3 @@ def __call__(self, waveform: np.ndarray, padding=160, chunk_length=None):
         log_spec = (log_spec + 4.0) / 4.0
 
         return log_spec
-
-    if win_length <= 0 or win_length > n_fft:
-        raise ValueError(
-            f"stft: expected 0 < win_length <= n_fft, but got win_length={win_length}"
-        )
-
-    if window is not None:
-        if window.ndim != 1 or window.shape[0] != win_length:
-            raise ValueError(
-                f"stft: expected a 1D window tensor of size equal to win_length={win_length}, "
-                f"but got window with size {window.shape}"
-            )
-
-    # Handle padding of the window if necessary
-    if win_length < n_fft:
-        left = (n_fft - win_length) // 2
-        window_ = np.zeros(n_fft, dtype=window.dtype)
-        window_[left : left + win_length] = window
-    else:
-        window_ = window
-
-    # Calculate the number of frames
-    n_frames = 1 + (length - n_fft) // hop_length
-
-    # Time to columns
-    input_tensor = np.lib.stride_tricks.as_strided(
-        input_tensor,
-        (batch, n_frames, n_fft),
-        (
-            input_tensor.strides[0],
-            hop_length * input_tensor.strides[1],
-            input_tensor.strides[1],
-        ),
-    )
-
-    if window_ is not None:
-        input_tensor = input_tensor * window_
-
-    # FFT and transpose
-    complex_fft = input_is_complex
-    onesided = onesided if onesided is not None else not complex_fft
-
-    if normalized:
-        norm = "ortho"
-    else:
-        norm = None
-
-    if complex_fft:
-        if onesided:
-            raise ValueError(
-                "Cannot have onesided output if window or input is complex"
-            )
-        output = np.fft.fft(input_tensor, n=n_fft, axis=-1, norm=norm)
-    else:
-        output = np.fft.rfft(input_tensor, n=n_fft, axis=-1, norm=norm)
-
-    output = output.transpose((0, 2, 1))
-
-    if input_tensor_1d:
-        output = output.squeeze(0)
-
-    return output if return_complex else np.real(output)