license: cc-by-nc-sa-4.0
pipeline_tag: fill-mask
language: en
tags:
- long_documents
datasets:
- wikipedia
model-index:
- name: kiddothe2b/hierarchical-transformer-LC1-mini-1024
results: []
Hierarchical Attention Transformer (HAT) / hierarchical-transformer-LC1-mini-1024
Model description
This is a Hierarchical Attention Transformer (HAT) model as presented in An Exploration of Hierarchical Attention Transformers for Efficient Long Document Classification (Chalkidis et al., 2022).
The model has been warm-started re-using the weights of miniature BERT (Turc et al., 2019), and continued pre-trained for MLM following the paradigm of Longformer released by Beltagy et al. (2020). It supports sequences of length up to 1,024.
HAT use a hierarchical attention, which is a combination of segment-wise and cross-segment attention operations. You can think segments as paragraphs or sentences.
Intended uses & limitations
You can use the raw model for masked language modeling, but it's mostly intended to be fine-tuned on a downstream task. See the model hub to look for fine-tuned versions on a task that interests you.
Note that this model is primarily aimed at being fine-tuned on tasks that use the whole document to make decisions, such as document classification, sequential sentence classification or question answering.
How to use
You can use this model directly with a pipeline for masked language modeling:
from transformers import pipeline
mlm_model = pipeline('fill-mask', model='kiddothe2b/hierarchical-transformer-LC1-mini-1024', trust_remote_code=True)
mlm_model("Hello I'm a <mask> model.")
You can also fine-tun it for SequenceClassification, SequentialSentenceClassification, and MultipleChoice down-stream tasks:
from transformers import AutoTokenizer, AutoModelforSequenceClassification
tokenizer = AutoTokenizer.from_pretrained("kiddothe2b/hierarchical-transformer-LC1-mini-1024", trust_remote_code=True)
doc_classifier = AutoModelforSequenceClassification(model='kiddothe2b/hierarchical-transformer-LC1-mini-1024', trust_remote_code=True)
Limitations and bias
The training data used for this model contains a lot of unfiltered content from the internet, which is far from neutral. Therefore, the model can have biased predictions.
Training procedure
Training and evaluation data
The model has been warm-started from google/bert_uncased_L-6_H-256_A-4 checkpoint and has been continued pre-trained for additional 50k steps on English Wikipedia.
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- distributed_type: tpu
- num_devices: 8
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- total_eval_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- training_steps: 50000
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 2.3959 | 0.2 | 10000 | 2.2258 |
| 2.3395 | 0.4 | 20000 | 2.1738 |
| 2.3082 | 0.6 | 30000 | 2.1404 |
| 2.273 | 0.8 | 40000 | 2.1145 |
| 2.262 | 1.14 | 50000 | 2.1004 |
Framework versions
- Transformers 4.19.0.dev0
- Pytorch 1.11.0+cu102
- Datasets 2.0.0
- Tokenizers 0.11.6
##Citing If you use HAT in your research, please cite An Exploration of Hierarchical Attention Transformers for Efficient Long Document Classification
@misc{chalkidis-etal-2022-hat,
url = {https://arxiv.org/abs/xxx},
author = {Chalkidis, Ilias and Dai, Xiang and Fergadiotis, Manos and Malakasiotis, Prodromos and Elliott, Desmond},
title = {An Exploration of Hierarchical Attention Transformers for Efficient Long Document Classification},
publisher = {arXiv},
year = {2022},
}