joshdevins's picture
New training version with all tokens labelled
def1aef unverified
|
raw
history blame
677 Bytes
metadata
language: en
license: apache-2.0
datasets:
  - conll2003

DistilBERT cased, fine-tuned for NER using the conll03 english dataset.

Versions

Transformers version: 4.3.1 Datasets version: 1.3.0

Training

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

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