MODEL.md

license	base_model	library_name	tags	model-index
apache-2.0	base_model_name	transformers	relation extraction nlp	name results iter-{dataset}-{model_name} task dataset metrics type relation-extraction name type dataset dataset name type value F1 f1 f1

ITER: Iterative Transformer-based Entity Recognition and Relation Extraction

This model checkpoint is part of the collection of models published alongside our paper ITER, accepted at EMNLP 2024.
To ease reproducibility and enable open research, our source code has been published on GitHub.

This model achieved an F1 score of {f1} on dataset {dataset}

Using ITER in your code

First, install ITER in your preferred environment:

pip install git+https://github.com/fleonce/iter

To use our model, refer to the following code:

from iter import {model_class}

model = {model_class}.from_pretrained("fleonce/iter-{dataset}-{model_name}")
tokenizer = model.tokenizer

encodings = tokenizer(
  "An art exhibit at the Hakawati Theatre in Arab east Jerusalem was a series of portraits of Palestinians killed in the rebellion .",
  return_tensors="pt"
)

generation_output = model.generate(
    encodings["input_ids"],
    attention_mask=encodings["attention_mask"],
)

# entities
print(generation_output.entities)

# relations between entities
print(generation_output.links)

Checkpoints

We publish checkpoints for the models performing best on the following datasets:

• ACE05:
  1. fleonce/iter-ace05-flant5-large
  2. fleonce/iter-ace05-flant5-xl
  3. fleonce/iter-ace05-deberta-large
• CoNLL04:
  1. fleonce/iter-conll04-flant5-large
  2. fleonce/iter-conll04-flant5-xl
  3. fleonce/iter-conll04-deberta-large
• ADE:
  1. fleonce/iter-ade-flant5-large
  2. fleonce/iter-ade-flant5-xl
  3. fleonce/iter-ade-deberta-large
• SciERC:
  1. fleonce/iter-scierc-flant5-large
  2. fleonce/iter-scierc-flant5-xl
  3. fleonce/iter-scierc-deberta-large
  4. fleonce/iter-scierc-scideberta-full
• CoNLL03:
  1. fleonce/iter-conll03-flant5-large
  2. fleonce/iter-conll03-flant5-xl
  3. fleonce/iter-conll03-deberta-large
• GENIA:
  1. fleonce/iter-genia-flant5-large
  2. fleonce/iter-genia-flant5-xl
  3. fleonce/iter-genia-deberta-large

Reproducibility

For each dataset, we selected the best performing checkpoint out of the 5 training runs we performed during training. This model was trained with the following hyperparameters:

• Seed: {seed}
• Config: {config}
• PyTorch {torch.version} with CUDA {torch.cuda} and precision {torch.precision}
• GPU: 1 {gpu}

Varying GPU and CUDA version as well as training precision did result in slightly different end results in our tests for reproducibility.

To train this model, refer to the following command:

python3 train.py --dataset {config} --transformer {base_model_name}{precision_command} --seed {seed}

@inproceedings{citation}