We use Flair for fine-tuning NER models on
HIPE-2022 datasets from
HIPE-2022 Shared Task.
All models are fine-tuned on A10 (24GB) and A100 (40GB) instances from
Lambda Cloud using Flair:
$ git clone
https://github.com/flairNLP/flair.git
$ cd flair && git checkout 419f13a05d6b36b2a42dd73a551dc3ba679f820c
$ pip3 install -e .
$ cd ..
Clone this repo for fine-tuning NER models:
$ git clone
https://github.com/stefan-it/hmTEAMS.git
$ cd hmTEAMS/bench
Authorize via Hugging… See the full description on the dataset page:
https://huggingface.co/datasets/stefan-it/autotrain-flair-hipe2022-fr-hmbert.
