huggingface/pytorch-image-models · error · ValueError

Invalid learning rate: {lr}

Error message

Invalid learning rate: {lr}

What it means

NAdamW optimizer constructor validation: learning rate must satisfy 0.0 <= lr. Negative values are rejected when timm.optim.NAdamW is created.

Source

Thrown at timm/optim/nadamw.py:56

        caution: enable caution
        corrected_weight_decay: apply corrected weight decay (lr**2 / max_lr)
    """

    def __init__(
            self,
            params: ParamsT,
            lr: float = 1e-3,
            betas: Tuple[float, float] = (0.9, 0.999),
            eps: float = 1e-8,
            weight_decay: float = 1e-2,
            caution: bool = False,
            corrected_weight_decay: bool = False,
            maximize: bool = False,
            foreach: Optional[bool] = None,
            capturable: bool = False,
    ):
        if not 0.0 <= lr:
            raise ValueError(f'Invalid learning rate: {lr}')
        if not 0.0 <= eps:
            raise ValueError(f'Invalid epsilon value: {eps}')
        if not 0.0 <= betas[0] < 1.0:
            raise ValueError(f'Invalid beta parameter at index 0: {betas[0]}')
        if not 0.0 <= betas[1] < 1.0:
            raise ValueError(f'Invalid beta parameter at index 1: {betas[1]}')
        if not 0.0 <= weight_decay:
            raise ValueError(f'Invalid weight_decay value: {weight_decay}')
        defaults = dict(
            lr=lr,
            betas=betas,
            eps=eps,
            weight_decay=weight_decay,
            caution=caution,
            corrected_weight_decay=corrected_weight_decay,
            foreach=foreach,
            maximize=maximize,
            capturable=capturable,

View on GitHub (pinned to 9a5261e31b)

Solutions

  1. Set lr to a valid non-negative value
  2. Validate hyperparameters before optimizer creation when they come from search/sweeps

Example fix

# before
opt = NAdamW(model.parameters(), lr=-1e-4)

# after
opt = NAdamW(model.parameters(), lr=1e-4)
Defensive patterns

Strategy: validation

Validate before calling

assert lr >= 0.0, f'lr must be >= 0, got {lr}'

Prevention

When it happens

Trigger: Calling timm.optim.NAdamW(params, lr=-0.001) or with lr computed negative/NaN from hyperparameter search.

Common situations: Sweep scripts producing negative lr, config typos, or lr read from an env var with a leading dash.

Related errors

AI-assisted analysis of huggingface/pytorch-image-models@9a5261e31b (2026-08-27). Data as JSON: /api/errors/8350fd6baf53e93c. Report an issue: GitHub.