Skip to content

[MathNLP 2022] 案牍术 natural language to code

License

Notifications You must be signed in to change notification settings

VisualJoyce/AnDuShu

Folders and files

NameName
Last commit message
Last commit date

Latest commit

3f784d3 · Sep 15, 2022

History

83 Commits
Sep 15, 2022
Nov 21, 2021
Sep 15, 2022
Feb 14, 2021
Oct 27, 2021
Jan 16, 2022
Dec 16, 2021
Apr 10, 2021
Dec 16, 2021
Mar 12, 2021
Dec 16, 2021
Dec 16, 2021
Sep 14, 2022

Repository files navigation

AuDuShu

AnDuShu

An inclusive natural language to code library built over allennlp 2.


This repository is using code from the following resources:

In this repo, we adapt the code to the latest allennlp version.


ATIS

PYTHONPATH=src ANNOTATION_DIR=$PWD/data/annotations/atis/ \
  allennlp train configs/atis/seq2seq/defaults.jsonnet \
  -s data/output/atis/seq2seq --include-package andushu
PYTHONPATH=src ANNOTATION_DIR=$PWD/data/annotations/atis/ \
  allennlp predict \
  --output-file=data/output/atis/seq2seq/seq2seq.jsonl \
  --predictor seq2seq \
  --include-package andushu \
  data/output/atis/seq2seq/model.tar.gz \
  data/annotations/atis/atis_test.jsonl 

GEOQUERY

PYTHONPATH=src ANNOTATION_DIR=$PWD/data/annotations/geoquery/ \
  allennlp train configs/geoquery/seq2seq/defaults.jsonnet \
  -s data/output/geoquery/seq2seq --include-package andushu
PYTHONPATH=src ANNOTATION_DIR=$PWD/data/annotations/geoquery/ \
  allennlp predict \
  --output-file=data/output/geoquery/seq2seq/seq2seq.jsonl \
  --predictor seq2seq \
  --include-package andushu data/output/geoquery/seq2seq/model.tar.gz \
  data/annotations/geoquery/geo_test.jsonl 

Math Word Problems

This repo includes the code for

@article{Tan2021InvestigatingMW,
  title={Investigating Math Word Problems using Pretrained Multilingual Language Models},
  author={Minghuan Tan and Lei Wang and Lingxiao Jiang and Jing Jiang},
  journal={ArXiv},
  year={2021},
  volume={abs/2105.08928}
}

Annotations

Annotations used for this paper can be found at

9B9A5987B7C1CF1482CAAF28B32AC0DE

Training

Training Math23K using bert-base-multilingual-cased.

CUDA_VISIBLE_DEVICES=2 PROJECT=math2tree SUB_PROJECT=math23k CONFIG=transformer_vocab SPACY_LANGUAGE=zh MODEL_NAME=bert-base-multilingual-cased OP_TYPE=disallow_pow bash docker_train.sh

Training over MathQA-Adapted without Pow.

CUDA_VISIBLE_DEVICES=2 PROJECT=math2tree SUB_PROJECT=mathqa CONFIG=transformer_vocab SPACY_LANGUAGE=zh MODEL_NAME=bert-base-multilingual-cased OP_TYPE=disallow_pow bash docker_train.sh

Training over MathQA-Adapted with Pow.

CUDA_VISIBLE_DEVICES=2 PROJECT=math2tree SUB_PROJECT=mathqa CONFIG=transformer_vocab SPACY_LANGUAGE=zh MODEL_NAME=bert-base-multilingual-cased OP_TYPE=allow_pow bash docker_train.sh

Training over MathXLing without Pow.

CUDA_VISIBLE_DEVICES=2 PROJECT=math2tree SUB_PROJECT=mathxling CONFIG=transformer_vocab SPACY_LANGUAGE=zh MODEL_NAME=bert-base-multilingual-cased OP_TYPE=disallow_pow bash docker_train.sh

Training over MathXLing with Pow.

CUDA_VISIBLE_DEVICES=2 PROJECT=math2tree SUB_PROJECT=mathxling CONFIG=transformer_vocab SPACY_LANGUAGE=zh MODEL_NAME=bert-base-multilingual-cased OP_TYPE=allow_pow bash docker_train.sh

Evaluation

CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/mathqa/transformer_vocab_en_disallow_pow_bert-base-multilingual-cased evaluate
CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/mathqa/transformer_vocab_en_allow_pow_bert-base-multilingual-cased evaluate

CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/mathqa/transformer_vocab_en_disallow_pow_bert-base-multilingual-cased evaluate 
CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/math23k/transformer_vocab_zh_disallow_pow_bert-base-multilingual-cased evaluate
 
CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/math23k/transformer_vocab_zh_allow_pow_bert-base-multilingual-cased evaluate
CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/math23k/transformer_vocab_en_allow_pow_bert-base-multilingual-cased evaluate

CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/math23k/transformer_vocab_en_disallow_pow_bert-base-multilingual-cased evaluate

CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/mathxling/transformer_vocab_zh_sallow_pow_bert-base-multilingual-cased evaluate
CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/mathxling/transformer_vocab_zh_allow_pow_bert-base-multilingual-cased evaluate

CUDA_VISIBLE_DEVICES=7 bash docker_eval.sh math2tree data/output/mathxling/transformer_vocab_zh_allow_pow_xlm-roberta-base evaluate