robustML/advertrain/dependencies/autoattack.py

• 4 F841: local variable 'counter' is assigned to but never used (and 3 similar)

  • Local variable 'counter' is assigned to but never used:

    393 counter = 0

  • Local variable 'n_reduced' is assigned to but never used:

    494 n_reduced = fl_oscillation.sum()

  • Local variable 'loss' is assigned to but never used:

    549 loss = -1e10 * torch.ones_like(acc).float()

  • Local variable 'eps_target' is assigned to but never used:

    625 eps_target = float(epss[-1])
• 9 E501: line too long (170 > 120 characters) (and 6 similar)

  • Line too long (170 > 120 characters):

    307 def attack_single_run(self, x: torch.Tensor, y: torch.Tensor, x_init: Optional[torch.Tensor] = None) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:

  • Line too long (163 > 120 characters):

    429 x_adv_1 = torch.clamp(x + self.normalize(x_adv_1 - x) * torch.min(self.eps * torch.ones_like(x).detach(), self.lp_norm(x_adv_1 - x)), 0.0, 1.0)

  • Line too long (163 > 120 characters):

    431 x_adv_1 = torch.clamp(x + self.normalize(x_adv_1 - x) * torch.min(self.eps * torch.ones_like(x).detach(), self.lp_norm(x_adv_1 - x)), 0.0, 1.0)

  • Line too long (170 > 120 characters):

    437 x_adv_1 = x_adv + step_size * sparsegrad.sign() / (sparsegrad.sign().abs().view(x.shape[0], -1).sum(dim=-1).view(-1, *[1]*(len(x.shape) - 1)) + 1e-10)

  • Line too long (153 > 120 characters):

    517 def perturb(self, x: torch.Tensor, y: Optional[torch.Tensor] = None, best_loss: bool = False, x_init: Optional[torch.Tensor] = None) -> torch.Tensor:

  • Line too long (147 > 120 characters):

    551 print('-------------------------- ', 'running {}-attack with epsilon {:.5f}'.format(self.norm, self.eps), '--------------------------')

  • Line too long (123 > 120 characters):

    561 print('using schedule [{}x{}]'.format('+'.join([str(c) for c in epss]), '+'.join([str(c) for c in iters])))

  • Line too long (172 > 120 characters):

    602 def decr_eps_pgd(self, x: torch.Tensor, y: torch.Tensor, epss: list, iters: list, use_rs: bool = True) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:

  • Line too long (121 > 120 characters):

    614 Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]: A tuple containing the final perturbed inputs,
• 5 E226: missing whitespace around arithmetic operator
  • 41 z = z.view(-1, *[1]*(len(x.shape) - 1))
  • 58 z = z.view(-1, *[1]*(len(x.shape) - 1))
  • 83 inu = 2*(indbs < u.shape[1]).float() - 1
  • 435 grad_topk = grad_topk[u, topk_curr].view(-1, *[1]*(len(x.shape) - 1))
  • 437 x_adv_1 = x_adv + step_size * sparsegrad.sign() / (sparsegrad.sign().abs().view(x.shape[0], -1).sum(dim=-1).view(-1, *[1]*(len(x.shape) - 1)) + 1e-10)
• 4 E714: test for object identity should be 'is not'
  • 209 assert not self.eps is None
  • 333 if not x_init is None:
  • 531 if not y is None and len(y.shape) == 0:
  • 634 if not x_init is None:
• 2 E128: continuation line under-indented for visual indent
  • 381 self.y_target)
  • 557 .4 * self.n_iter_orig]
• 2 E303: too many blank lines (2)
  • 201 def init_hyperparam(self, x: torch.Tensor) -> None:
  • 517 def perturb(self, x: torch.Tensor, y: Optional[torch.Tensor] = None, best_loss: bool = False, x_init: Optional[torch.Tensor] = None) -> torch.Tensor:
• 1 E722: do not use bare 'except'
  • 272 except:
• 1 E731: do not assign a lambda expression, use a def
  • 348 criterion_indiv = lambda x, y: -1. * F.cross_entropy(x, y, reduction='none')
• 1 E741: ambiguous variable name 'l'
  • 79 l = -torch.clone(y).abs()