huggingface/pytorch-image-models · error · ValueError

Invalid epsilon value: {}

Error message

Invalid epsilon value: {}

What it means

Nvnovograd optimizer constructor validation: epsilon must satisfy 0.0 <= eps. Negative epsilon would corrupt the gradient-normalization denominator and is rejected.

Source

Thrown at timm/optim/nvnovograd.py:45

        amsgrad (boolean, optional): whether to use the AMSGrad variant of this
            algorithm from the paper `On the Convergence of Adam and Beyond`_
            (default: False)
    """

    def __init__(
            self,
            params,
            lr=1e-3,
            betas=(0.95, 0.98),
            eps=1e-8,
            weight_decay=0,
            grad_averaging=False,
            amsgrad=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]))
        defaults = dict(
            lr=lr,
            betas=betas,
            eps=eps,
            weight_decay=weight_decay,
            grad_averaging=grad_averaging,
            amsgrad=amsgrad,
        )

        super(NvNovoGrad, self).__init__(params, defaults)

    def __setstate__(self, state):
        super(NvNovoGrad, self).__setstate__(state)
        for group in self.param_groups:

View on GitHub (pinned to 9a5261e31b)

Solutions

  1. Use a positive eps like 1e-8 (default is 1e-8)
  2. Omit eps to accept the default

Example fix

# before
opt = Nvnovograd(model.parameters(), eps=-1e-8)

# after
opt = Nvnovograd(model.parameters(), eps=1e-8)
Defensive patterns

Strategy: validation

Validate before calling

assert eps >= 0.0

Prevention

When it happens

Trigger: Calling timm.optim.Nvnovograd(params, eps=-1e-8).

Common situations: Config typos or copy-paste of a negative numerical-stability constant.

Related errors

AI-assisted analysis of huggingface/pytorch-image-models@9a5261e31b (2026-08-27). Data as JSON: /api/errors/4d4bfc41b3dae760. Report an issue: GitHub.