Views
No views yet
Yale-LILY/reastap-large-finetuned-wtq is initialized with Yale-LILY/reastap-large and finetuned on WikiTableQuestions.1from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
2import pandas as pd
3
4tokenizer = AutoTokenizer.from_pretrained("Yale-LILY/reastap-large-finetuned-wtq")
5model = AutoModelForSeq2SeqLM.from_pretrained("Yale-LILY/reastap-large-finetuned-wtq")
6
7data = {
8 "year": [1896, 1900, 1904, 2004, 2008, 2012],
9 "city": ["athens", "paris", "st. louis", "athens", "beijing", "london"]
10}
11table = pd.DataFrame.from_dict(data)
12
13query = "In which year did beijing host the Olympic Games?"
14encoding = tokenizer(table=table, query=query, return_tensors="pt")
15
16outputs = model.generate(**encoding)
17
18print(tokenizer.batch_decode(outputs, skip_special_tokens=True))
19# [' 2008']1@inproceedings{zhao-etal-2022-reastap,
2 title = "{R}eas{TAP}: Injecting Table Reasoning Skills During Pre-training via Synthetic Reasoning Examples",
3 author = "Zhao, Yilun and
4 Nan, Linyong and
5 Qi, Zhenting and
6 Zhang, Rui and
7 Radev, Dragomir",
8 booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing",
9 month = dec,
10 year = "2022",
11 address = "Abu Dhabi, United Arab Emirates",
12 publisher = "Association for Computational Linguistics",
13 url = "https://aclanthology.org/2022.emnlp-main.615",
14 pages = "9006--9018",
15 abstract = "Reasoning over tabular data requires both table structure understanding and a broad set of table reasoning skills. Current models with table-specific architectures and pre-training methods perform well on understanding table structures, but they still struggle with tasks that require various table reasoning skills. In this work, we develop ReasTAP to show that high-level table reasoning skills can be injected into models during pre-training without a complex table-specific architecture design. We define 7 table reasoning skills, such as numerical operation, temporal comparison, and conjunction. Each reasoning skill is associated with one example generator, which synthesizes questions over semi-structured tables according to the sampled templates. We model the table pre-training task as a sequence generation task and pre-train ReasTAP to generate precise answers of the synthetic examples. ReasTAP is evaluated on four benchmarks covering three downstream tasks including 1) WikiSQL-Weak and WikiTQ for Table Question Answering, 2) TabFact for Table Fact Verification, and 3) LogicNLG for Faithful Table-to-Text Generation. Experimental results demonstrate that ReasTAP achieves new state-of-the-art results on all of them and delivers a significant improvement under low-resource setting. Our code is publicly available at https://github.com/Yale-LILY/ReasTAP.",
16}