deepfakes/faceswap · error · ValueError

Hex color codes should start with a '#' and be 6 characters

Error message

Hex color codes should start with a '#' and be 6 characters long. Got: '{value}'

What it means

Apple Silicon GPU stats plugin (lib/gpu_stats/apple_silicon.py) sanity-checks the Torch MPS backend by calling torch.mps.driver_allocated_memory(). If that call raises RuntimeError, Faceswap wraps it as FaceswapError, since a broken MPS stack means no usable Apple GPU.

Source

Thrown at lib/config/objects.py:400

        if self.datatype is list:
            if not isinstance(value, (str, list)):
                raise ValueError(f"[{self._name}] List values should be set as a Str or List. Got "
                                 f"{type(value)} ({value})")
            value = cast(T, self._parse_list(value))

        if not isinstance(value, self.datatype):
            raise ValueError(
                f"[{self._name}] Expected {self.datatype} got {type(value)} ({value})")

        if isinstance(self.choices, list) and self.choices:
            assert isinstance(value, (list, str))
            value = cast(T, self._validate_selection(value))

        if self.choices == "colorchooser":
            assert isinstance(value, str)
            if not value.startswith("#") or len(value) != 7:
                raise ValueError(f"Hex color codes should start with a '#' and be 6 "
                                 f"characters long. Got: '{value}'")

        self._value = value

    def set_name(self, name: str) -> None:
        """Set the logging name for this object for display purposes

        Parameters
        ----------
        name
            The name to assign to this option
        """
        logger.debug("Setting name to '%s'", name)
        assert isinstance(name, str) and name
        self._name = name

    def __call__(self) -> T:
        """Obtain the currently stored configuration value

View on GitHub (pinned to f530cb7508)

Solutions

1. Verify MPS outside Faceswap: `python -c "import torch; print(torch.backends.mps.is_available()); print(torch.mps.driver_allocated_memory())"` and read the original error
2. Reinstall a torch build matching your macOS/arch (e.g. pip install --force-reinstall torch on native arm64, no Rosetta)
3. Update macOS/Xcode command line tools so Metal drivers match the torch build; as a fallback run with CPU (`--cpu`) while the GPU stack is repaired

Example fix

# before
python faceswap.py --apple-silicon train ...  # -> FaceswapError wrapping torch.mps RuntimeError

# after
# 1) verify the stack
python -c "import torch; print(torch.backends.mps.is_available())"
# 2) reinstall matching torch, then retry
pip uninstall -y torch && pip install torch
# 3) meanwhile run CPU-only
python faceswap.py --cpu train ...

import re

def valid_hex_color(v: str) -> bool:
    return bool(re.fullmatch(r"#[0-9a-fA-F]{6}", v))

assert valid_hex_color(option_value), f"bad color: {option_value!r}"

import re

def is_hex_color(v) -> bool:
    """True when v is a '#rrggbb' string accepted by colorchooser options."""
    return isinstance(v, str) and bool(re.fullmatch(r"#[0-9a-fA-F]{6}", v))

try:
    option.set(color)
except ValueError:
    r, g, b = tuple(int(color.lstrip("#")[i:i+2], 16) for i in (0, 2, 4))
    option.set("#%02x%02x%02x" % (r, g, b))

When it happens

Trigger: Selecting apple Silicon GPU stats (e.g. `faceswap.py --apple-silicon` or auto-detection on an M-series Mac) with a torch build lacking MPS support, a too-old macOS/torch version, or a corrupted Metal/torch install; also when torch imports but the MPS device is unavailable.

Common situations: Upgrading/downgrading torch or macOS and breaking MPS; conda envs shipping CPU-only torch on ARM; running under Rosetta where MPS is unavailable; beta macOS with incompatible Metal drivers.

Related errors

• An unhandled exception occurred initializing the device via
• Dataclass params {sorted(required)} should be a subset of di
• There was an error reading from the Nvidia Machine Learning
• You do not have enough GPU memory available to train the sel
• {arch}' is not compatible with your version of Keras. The mi

AI-assisted analysis of deepfakes/faceswap@f530cb7508 (2026-08-15). Data as JSON: /api/errors/ede30864f2fc7db8. Report an issue: GitHub.