lewtun's picture
lewtun HF staff
Add evaluation results on conll2003 dataset
d64628f
|
raw
history blame
1.8 kB
metadata
language: en
license: apache-2.0
datasets:
  - conll2003
model-index:
  - name: elastic/distilbert-base-cased-finetuned-conll03-english
    results:
      - task:
          type: token-classification
          name: Token Classification
        dataset:
          name: conll2003
          type: conll2003
          config: conll2003
          split: validation
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.9834432212868665
            verified: true
          - name: Precision
            type: precision
            value: 0.9857564461012737
            verified: true
          - name: Recall
            type: recall
            value: 0.9882123948925569
            verified: true
          - name: F1
            type: f1
            value: 0.9869828926905132
            verified: true
          - name: loss
            type: loss
            value: 0.07748260349035263
            verified: true

DistilBERT base cased, fine-tuned for NER using the conll03 english dataset. Note that this model is sensitive to capital letters — "english" is different than "English". For the case insensitive version, please use elastic/distilbert-base-uncased-finetuned-conll03-english.

Versions

  • Transformers version: 4.3.1
  • Datasets version: 1.3.0

Training

$ run_ner.py \
  --model_name_or_path distilbert-base-cased \
  --label_all_tokens True \
  --return_entity_level_metrics True \
  --dataset_name conll2003 \
  --output_dir /tmp/distilbert-base-cased-finetuned-conll03-english \
  --do_train \
  --do_eval

After training, we update the labels to match the NER specific labels from the dataset conll2003