metadata
license: gpl-3.0
tags:
- DocVQA
- Document Question Answering
- Document Visual Question Answering
datasets:
- rubentito/mp-docvqa
language:
- en
BigBird base (ITC) fine-tuned on MP-DocVQA
This is BigBird-base trained on TriviaQA from Google hub and fine-tuned on Multipage DocVQA (MP-DocVQA) dataset.
- Due to Huggingface implementation, the global tokens are defined according to the Internal Transformer Construction (ITC) strategy.
This model was used as a baseline in Hierarchical multimodal transformers for Multi-Page DocVQA.
- Results on the MP-DocVQA dataset are reported in Table 2.
- Training hyperparameters can be found in Table 8 of Appendix D.
How to use
How to use this model to perform inference on a sample question and context in PyTorch:
from transformers import BigBirdForQuestionAnswering, BigBirdTokenizerFast
# by default its in `block_sparse` mode with num_random_blocks=3, block_size=64
model = BigBirdForQuestionAnswering.from_pretrained("rubentito/bigbird-base-itc-mpdocvqa")
# you can change `attention_type` to full attention like this:
model = BigBirdForQuestionAnswering.from_pretrained("rubentito/bigbird-base-itc-mpdocvqa", attention_type="original_full")
# you can change `block_size` & `num_random_blocks` like this:
model = BigBirdForQuestionAnswering.from_pretrained("rubentito/bigbird-base-itc-mpdocvqa", block_size=16, num_random_blocks=2)
tokenizer = BigBirdTokenizerFast.from_pretrained("rubentito/bigbird-base-itc-mpdocvqa")
question = "Replace me by any text you'd like."
context = "Put some context for answering"
encoded_input = tokenizer(question, context, return_tensors='pt')
output = model(**encoded_input)
start_pos = torch.argmax(output.start_logits, dim=-1).item()
end_pos = torch.argmax(output.end_logits, dim=-1).item()
context_tokens = tokenizer.convert_ids_to_tokens(encoded_input["input_ids"][0].tolist())
answer_tokens = context_tokens[start_pos: end_pos]
answer = tokenizer.decode(tokenizer.convert_tokens_to_ids(answer_tokens))
Model results
Extended experimentation can be found in Table 2 of Hierarchical multimodal transformers for Multi-Page DocVQA. You can also check the live leaderboard at the RRC Portal.
| Model | HF name | Parameters | ANLS | APPA |
|---|---|---|---|---|
| Bert large | rubentito/bert-large-mpdocvqa | 334M | 0.4183 | 51.6177 |
| Longformer base | rubentito/longformer-base-mpdocvqa | 148M | 0.5287 | 71.1696 |
| BigBird ITC base | rubentito/bigbird-base-itc-mpdocvqa | 131M | 0.4929 | 67.5433 |
| LayoutLMv3 base | rubentito/layoutlmv3-base-mpdocvqa | 125M | 0.4538 | 51.9426 |
| T5 base | rubentito/t5-base-mpdocvqa | 223M | 0.5050 | 0.0000 |
| Hi-VT5 | TBA | 316M | 0.6201 | 79.23 |
Citation Information
@article{tito2022hierarchical,
title={Hierarchical multimodal transformers for Multi-Page DocVQA},
author={Tito, Rub{\`e}n and Karatzas, Dimosthenis and Valveny, Ernest},
journal={arXiv preprint arXiv:2212.05935},
year={2022}
}