models/initializer.py

• 20 E501: line too long (117 > 79 characters) (and 15 similar)

  • Line too long (117 > 79 characters):

    10 def initialize_model(config, d_out, is_featurizer=False, full_dataset=None, is_pooled=True, include_projector=False):

  • Line too long (122 > 79 characters):

    16 - is_featurizer (bool): whether to return a model or a (featurizer, classifier) pair that constitutes a model.

  • Line too long (113 > 79 characters):

    19 - featurizer: a model that outputs feature Tensors of shape (batch_size, ..., feature dimensionality)

  • Line too long (127 > 79 characters):

    20 - classifier: a model that takes in feature Tensors and outputs predictions. In most cases, this is a linear layer.

  • Line too long (88 > 79 characters):

    22 - model: a model that is equivalent to nn.Sequential(featurizer, classifier)

  • Line too long (104 > 79 characters):

    28 featurizer = ThreeWLGNNNet(gnn_type=config.model, num_tasks=None, **config.model_kwargs)

  • Line too long (100 > 79 characters):

    32 model = ThreeWLGNNNet(gnn_type=config.model, num_tasks=d_out, **config.model_kwargs)

  • Line too long (122 > 79 characters):

    34 assert config.algorithm == 'ERM' or config.algorithm == 'IRM' # combinations with other algorithms not checked

  • Line too long (94 > 79 characters):

    36 featurizer = MLP(gnn_type=config.model, num_tasks=None, **config.model_kwargs)

  • Line too long (90 > 79 characters):

    40 model = MLP(gnn_type=config.model, num_tasks=d_out, **config.model_kwargs)

  • Line too long (119 > 79 characters):

    44 featurizer = GNN(gnn_type=config.model, num_tasks=None, is_pooled=is_pooled, **config.model_kwargs)

  • Line too long (119 > 79 characters):

    48 featurizer = GNN(gnn_type=config.model, num_tasks=None, is_pooled=is_pooled, **config.model_kwargs)

  • Line too long (111 > 79 characters):

    53 model = GNN(gnn_type=config.model, num_tasks=d_out, is_pooled=is_pooled, **config.model_kwargs)

  • Line too long (81 > 79 characters):

    63 in_edge_features=full_dataset.num_edge_features,

  • Line too long (82 > 79 characters):

    64 d_in_node_encoder=full_dataset.d_in_node_encoder,

  • Line too long (82 > 79 characters):

    65 d_in_edge_encoder=full_dataset.d_in_edge_encoder,

  • Line too long (84 > 79 characters):

    67 dataset_group=config.model_kwargs['dataset_group'])

  • Line too long (93 > 79 characters):

    92 'not (featurizer, pooler, classifier), i.e. whole graph embeddings'

  • Line too long (162 > 79 characters):

    94 projector = nn.Sequential(nn.Linear(graph_embedding_dim, graph_embedding_dim), nn.ReLU(inplace=True), nn.Linear(graph_embedding_dim, graph_embedding_dim))

  • Line too long (90 > 79 characters):

    96 model = (featurizer, projector, classifier, nn.Sequential(featurizer, classifier))
• 3 E203: whitespace before ':'
  • 33 elif config.model == 'mlp' :
  • 43 if is_pooled :
  • 47 else :
• 2 W291: trailing whitespace
  • 84 # Usage of (featurizer, projector, classifier) in
  • 86 if include_projector:
• 1 E261: at least two spaces before inline comment
  • 34 assert config.algorithm == 'ERM' or config.algorithm == 'IRM' # combinations with other algorithms not checked
• 1 W293: blank line contains whitespace
  • 97