algorithms/MLDG.py

• 1 F401: 'torch.nn.functional as F' imported but unused
  • 2 import torch.nn.functional as F
• 6 E203: whitespace before ':'
  • 53 for i in range(rand_group_index.shape[0]) :
  • 74 try :
  • 85 except :
  • 97 try :
  • 99 except :
  • 106 if grad_inner_j is not None :
• 6 E501: line too long (92 > 79 characters) (and 5 similar)

  • Line too long (92 > 79 characters):

    17 model = initialize_model(config, d_out=d_out, is_featurizer=False).to(config.device)

  • Line too long (80 > 79 characters):

    56 group_pairs.append((group_indices[group_i], group_indices[group_j]))

  • Line too long (82 > 79 characters):

    84 p_tgt.grad.data.add_(p_src.grad.data / n_groups_per_batch)

  • Line too long (107 > 79 characters):

    98 grad_inner_j = torch.autograd.grad(loss_inner_j, inner_net.parameters(), allow_unused=True)

  • Line too long (95 > 79 characters):

    109 p.grad.data.add_(self.config.mldg_beta * g_j.data / n_groups_per_batch)

  • Line too long (89 > 79 characters):

    130 return self.loss.compute(results['y_pred'], results['y_true'], return_dict=False)
• 2 E722: do not use bare 'except'
  • 85 except :
  • 99 except :
• 1 E302: expected 2 blank lines, found 1
  • 10 class MLDG(SingleModelAlgorithm):
• 1 E303: too many blank lines (2)
  • 31 def update(self, batch):