TypeError: round_f1() takes 2 positional arguments but 3 were given

1

I am trying to create binary classifier using BERT’s RuBert.
Here is my config file:

model = {
  "dataset_reader": {
    "class_name": "basic_classification_reader",
    "x": "title",
    "y": "label",
    "data_path": "/content/drive/MyDrive/AI/datasets/headers_news/splitted/",
    "train": "train.csv",
    "test": "test.csv",
    "valid": "validation.csv"
  },
  "dataset_iterator": {
    "class_name": "basic_classification_iterator",
    "seed": 42,
    "shuffle": True
  },
  "chainer": {
    "in": [
      "x"
    ],
    "in_y": [
      "y"
    ],
    "pipe": [
      {
        "class_name": "bert_preprocessor",
        "vocab_file": "/content/drive/MyDrive/AI/config/rubert_cased_L-12_H-768_A-12_v2/vocab.txt",
        "do_lower_case": False,
        "max_seq_length": 512,
        "in": [
          "x"
        ],
        "out": [
          "bert_features"
        ]
      },
      {
        "id": "classes_vocab",
        "class_name": "simple_vocab",
        "fit_on": [
          "y"
        ],
        "save_path": "bert-news-classifier/data/binary_classes.dict",
        "load_path": "bert-news-classifier/data/binary_classes.dict",
        "in": "y",
        "out": "y_ids"
      },
      {
        "in": "y_ids",
        "out": "y_onehot",
        "class_name": "one_hotter",
        "depth": 2,
        "single_vector": True
      },
      {
        "class_name": "bert_classifier",
        "n_classes": 2,
        "return_probas": True,
        "one_hot_labels": True,
        "bert_config_file": "/content/drive/MyDrive/AI/config/rubert_cased_L-12_H-768_A-12_v2/bert_config.json",
        "pretrained_bert": "/content/drive/MyDrive/AI/config/rubert_cased_L-12_H-768_A-12_v2/bert_model.ckpt",
        "save_path": "bert-news-classifier/data/bert_model/model",
        "load_path": "bert-news-classifier/data/bert_model/model",
        "learning_rate": 2e-6,
        "keep_prob": 0.5,
        "load_before_drop": True,
        "loss": "cross-entropy",
        "in": [
          "bert_features"
        ],
        "in_y": [
          "y_onehot"
        ],
        "out": [
          "y_pred_probas"
        ]
      },
      {
        "in": "y_pred_probas",
        "out": "y_pred_ids",
        "class_name": "proba2labels",
        "max_proba": True
      },
      {
        "in": "y_pred_ids",
        "out": "y_pred_labels",
        "ref": "classes_vocab"
      }
    ],
    "out": [
      "y_pred_labels", "y_pred_probas"
    ]
  },
  "train": {
    "batch_size": 8,
    "epochs": 5,
    "metrics": [
    "f1",
    {
      "name": "accuracy",
      "inputs": ["y_onehot", "y_pred_labels"]
    },
    {
      "name": "roc_auc",
      "inputs": ["y_onehot", "y_pred_probas"]
    }
    ],
    "show_examples": False,
    "pytest_max_batches": 2,
    "validation_patience": 2,
    "val_every_n_epochs": 1,
    "log_every_n_epochs": 1,
    "tensorboard_log_dir": "bert-news-classifier/data/logs/",
    "evaluation_targets": [
      "train",
      "valid",
    ],
    "class_name": "nn_trainer"
  },
  "metadata": {
    "variables": {
      "ROOT_PATH": "~/.deeppavlov",
      "DOWNLOADS_PATH": "{ROOT_PATH}/downloads",
      "MODELS_PATH": "{ROOT_PATH}/models",
      "MODEL_PATH": "{MODELS_PATH}/classifiers/rusentiment_bert_v0/"
    },
    "requirements": [
      "deeppavlov/requirements/tf.txt",
      "deeppavlov/requirements/bert_dp.txt"
    ],
    "download": [
      {
        "url": "http://files.deeppavlov.ai/deeppavlov_data/bert/rubert_cased_L-12_H-768_A-12_v2.tar.gz",
        "subdir": "/content/bert_models"
      },
      {
        "url": "http://files.deeppavlov.ai/deeppavlov_data/classifiers/rubert_cased_L-12_H-768_A-12_v2.tar.gz",
        "subdir": "drive/MyDrive/bert_model/model_mix_73/classifier"
      }
    ]
  }
}

but I am getting the following error:

TypeError: round_f1() takes 2 positional arguments but 3 were given

The structure of the dataset has no problem. It has 2 columns (text and label)
What am I doing wrong?

2

Hi!
In DeepPavlov out + in_y is fed to metrics by default. In your case out:

"out": [
      "y_pred_labels", "y_pred_probas"
    ]

and in_y:

"in_y": [
      "y"
    ],

There are three values in total in out + in_y. F-1 measure expects only two values.
So, you need to specify inputs to F-1 metric in the similar way as for accuracy and roc_auc in your configuration file.