algorithms/groupDRO.py

• 8 E501: line too long (142 > 79 characters) (and 7 similar)

  • Line too long (142 > 79 characters):

    12 title={Distributionally robust neural networks for group shifts: On the importance of regularization for worst-case generalization},

  • Line too long (96 > 79 characters):

    13 author={Sagawa, Shiori and Koh, Pang Wei and Hashimoto, Tatsunori B and Liang, Percy},

  • Line too long (95 > 79 characters):

    19 def __init__(self, config, d_out, grouper, loss, metric, n_train_steps, is_group_in_train):

  • Line too long (80 > 79 characters):

    63 Takes an output of SingleModelAlgorithm.process_batch() and computes the

  • Line too long (81 > 79 characters):

    64 optimized objective. For group DRO, the objective is the weighted average

  • Line too long (82 > 79 characters):

    67 - results (dictionary): output of SingleModelAlgorithm.process_batch()

  • Line too long (94 > 79 characters):

    81 Process the batch, update the log, and update the model, group weights, and scheduler.

  • Line too long (109 > 79 characters):

    101 self.group_weights = self.group_weights * torch.exp(self.group_weights_step_size * group_losses.data)
• 1 W291: trailing whitespace
  • 16 }