huggingface/pytorch-image-models · error · ValueError

Invalid beta parameter at index 0: {}

Error message

Invalid beta parameter at index 0: {}

What it means

Mars requires beta1 (betas[0]) to satisfy 0 <= beta1 < 1, the standard constraint for exponential-moving-average decay coefficients. Values outside this range produce non-convergent momentum estimates.

Source

Thrown at timm/optim/mars.py:117

            self,
            params: ParamsT,
            lr: float = 3e-3,
            betas: Tuple[float, float] = (0.9, 0.99),
            eps: float = 1e-8,
            weight_decay: float = 0.,
            gamma: float = 0.025,
            mars_type: str = "adamw",
            optimize_1d: bool = False,
            lr_1d_factor: float = 1.0,
            betas_1d: Optional[Tuple[float, float]] = None,
            caution: bool = False
    ):
        if not 0.0 <= lr:
            raise ValueError("Invalid learning rate: {}".format(lr))
        if not 0.0 <= eps:
            raise ValueError("Invalid epsilon value: {}".format(eps))
        if not 0.0 <= betas[0] < 1.0:
            raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0]))
        if not 0.0 <= betas[1] < 1.0:
            raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1]))
        assert mars_type in ["adamw", "lion"], "MARS type not supported"

        defaults = dict(
            lr=lr,
            betas=betas,
            eps=eps,
            weight_decay=weight_decay,
            mars_type=mars_type,
            gamma=gamma,
            optimize_1d=optimize_1d,
            lr_1d_factor=lr_1d_factor,
            betas_1d=betas_1d or betas,
            caution=caution,
        )
        super(Mars, self).__init__(params, defaults)

View on GitHub (pinned to 9a5261e31b)

Solutions

1. Use betas with beta1 in [0,1), e.g. (0.9, 0.999)
2. Verify you passed (beta1, beta2) in the right order
3. Check Mars-specific betas_1d if you also configure the 1D param group

Example fix

# before
opt = Mars(model.parameters(), betas=(1.0, 0.999))
# after
opt = Mars(model.parameters(), betas=(0.9, 0.999))

b1, b2 = cfg.betas
assert 0 <= b1 < 1 and 0 <= b2 < 1, 'betas must be in [0,1)'

def are_valid_betas(betas: tuple) -> bool:
    return (len(betas) == 2 and all(isinstance(b, (int, float)) and 0 <= b < 1 for b in betas))

When it happens

Trigger: Calling Mars(params, betas=(1.0, 0.999)) or betas=(-0.1, 0.999); beta1=1 is explicitly rejected.

Common situations: Copying betas from a config where beta1 was set to 1 for 'full momentum'; ordering mixups passing (beta2, beta1).

Related errors

• Invalid beta parameter at index 1: {}
• Invalid learning rate: {}
• Invalid epsilon value: {}
• Momentum {momentum} must be in the range [0,1]
• Learning rate {lr} must be positive

AI-assisted analysis of huggingface/pytorch-image-models@9a5261e31b (2026-08-27). Data as JSON: /api/errors/56dc30577a71c6f6. Report an issue: GitHub.