amazon-science/tanl amazon-science/tanl

  • Stars
    star
    113
  • Rank 310,115 (Top 7 %)
  • Language
    Python
  • License
    Apache License 2.0
  • Created over 3 years ago
  • Updated over 1 year ago
Twitter logo Share LinkedIn logo Share Facebook logo Share
amazon-science/tanl
github

amazon-science

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Structured Prediction as Translation between Augmented Natural Languages

TANL: Structured Prediction as Translation between Augmented Natural Languages

Code for the paper "Structured Prediction as Translation between Augmented Natural Languages" (ICLR 2021) and fine-tuned multi-task model.

If you use this code, please cite the paper using the bibtex reference below.

@inproceedings{tanl,
    title={Structured Prediction as Translation between Augmented Natural Languages},
    author={Giovanni Paolini and Ben Athiwaratkun and Jason Krone and Jie Ma and Alessandro Achille and Rishita Anubhai and Cicero Nogueira dos Santos and Bing Xiang and Stefano Soatto},
    booktitle={9th International Conference on Learning Representations, {ICLR} 2021},
    year={2021},
}

Requirements

  • Python 3.6+
  • PyTorch (tested with version 1.7.1)
  • Transformers (tested with version 4.0.0)
  • NetworkX (tested with version 2.5, only used in coreference resolution)

You can install all required Python packages with pip install -r requirements.txt

Datasets

By default, datasets are expected to be in data/DATASET_NAME. Dataset-specific code is in datasets.py.

The CoNLL04 and ADE datasets (joint entity and relation extraction) in the correct format can be downloaded using https://github.com/markus-eberts/spert/blob/master/scripts/fetch_datasets.sh. For other datasets, we provide sample processing code which does not necessarily match the format of publicly available versions (we do not plan to adapt the code to load datasets in other formats).

Running the code

Use the following command: python run.py JOB

The JOB argument refers to a section of the config file, which by default is config.ini. A sample config file is provided, with settings that allow for a faster training and less memory usage than the settings used to obtain the final results in the paper.

For example, to replicate the paper's results on CoNLL04, have the following section in the config file:

[conll04_final]
datasets = conll04
model_name_or_path = t5-base
num_train_epochs = 200
max_seq_length = 256
max_seq_length_eval = 512
train_split = train,dev
per_device_train_batch_size = 8
per_device_eval_batch_size = 16
do_train = True
do_eval = False
do_predict = True
episodes = 1-10
num_beams = 8

Then run python run.py conll04_final. Note that the final results will differ slightly from the ones reported in the paper, due to small code changes and randomness.

Config arguments can be overwritten by command line arguments. For example: python run.py conll04_final --num_train_epochs 50.

Additional details

If do_train = True, the model is trained on the given train split (e.g., 'train') of the given datasets. The final weights and intermediate checkpoints are written in a directory such as experiments/conll04_final-t5-base-ep200-len256-b8-train, with one subdirectory per episode. Results in JSON format are also going to be saved there.

In every episode, the model is trained on a different (random) permutation of the training set. The random seed is given by the episode number, so that every episode always produces the same exact model.

Once a model is trained, it is possible to evaluate it without training again. For this, set do_train = False or (more easily) provide the -e command-line argument: python run.py conll04_final -e.

If do_eval = True, the model is evaluated on the 'dev' split. If do_predict = True, the model is evaluated on the 'test' split.

Arguments

The following are the most important command-line arguments for the run.py script. Run python run.py -h for the full list.

  • -c CONFIG_FILE: specify config file to use (default is config.ini)
  • -e: only run evaluation (overwrites the setting do_train in the config file)
  • -a: evaluate also intermediate checkpoints, in addition to the final model
  • -v : print results for each evaluation run
  • -g GPU: specify which GPU to use for evaluation

The following are the most important arguments for the config file. See the sample config file to understand the format.

  • datasets (str): comma-separated list of datasets for training
  • eval_datasets (str): comma-separated list of datasets for evaluation (default is the same as for training)
  • model_name_or_path (str): path to pretrained model or model identifier from huggingface.co/models (e.g. t5-base)
  • do_train (bool): whether to run training (default is False)
  • do_eval (bool): whether to run evaluation on the dev set (default is False)
  • do_predict (bool): whether to run evaluation on the test set (default is False)
  • train_split (str): comma-separated list of data splits for training (default is train)
  • num_train_epochs (int): number of train epochs
  • learning_rate (float): initial learning rate (default is 5e-4)
  • train_subset (float > 0 and <=1): portion of training data to effectively use during training (default is 1, i.e., use all training data)
  • per_device_train_batch_size (int): batch size per GPU during training (default is 8)
  • per_device_eval_batch_size (int): batch size during evaluation (default is 8; only one GPU is used for evaluation)
  • max_seq_length (int): maximum input sequence length after tokenization; longer sequences are truncated
  • max_output_seq_length (int): maximum output sequence length (default is max_seq_length)
  • max_seq_length_eval (int): maximum input sequence length for evaluation (default is max_seq_length)
  • max_output_seq_length_eval (int): maximum output sequence length for evaluation (default is max_output_seq_length or max_seq_length_eval or max_seq_length)
  • episodes (str): episodes to run (default is 0; an interval can be specified, such as 1-4; the episode number is used as the random seed)
  • num_beams (int): number of beams for beam search during generation (default is 1)
  • multitask (bool): if True, the name of the dataset is prepended to each input sentence (default is False)

See arguments.py and transformers.TrainingArguments for additional config arguments.

Fine-tuned multi-task model

The weights of our multi-task model (released under the CC BY 4.0 license) can be downloaded here: https://tanl.s3.amazonaws.com/tanl-multitask.zip

Extract the zip file in the experiments/ directory. This will create a subdirectory called multitask-t5-base-ep50-len512-b8-train,dev-overlap96. For example, to test the multi-task model on the CoNLL04 dataset, run python run.py multitask -e --eval_datasets conll04.

Note that: the multitask job is defined in config.ini; the -e flag is used to skip training and run evaluation only; the name of the subdirectory containing the weights is compatible with the definition of the multitask job.

The multi-task model was fine-tuned as described in the paper. The results differ slightly from what is reported in the paper due to small code changes.

Licenses

The code of this repository is released under the Apache 2.0 license. The weights of the fine-tuned multi-task model are released under the CC BY 4.0 license.

More Repositories

1

mm-cot

Official implementation for "Multimodal Chain-of-Thought Reasoning in Language Models" (stay tuned and more will be updated)
Python
3,727
star
2

chronos-forecasting

Chronos: Pretrained (Language) Models for Probabilistic Time Series Forecasting
Python
2,202
star
3

auto-cot

Official implementation for "Automatic Chain of Thought Prompting in Large Language Models" (stay tuned & more will be updated)
Jupyter Notebook
1,218
star
4

patchcore-inspection

Python
479
star
5

siam-mot

SiamMOT: Siamese Multi-Object Tracking
Python
458
star
6

alexa-teacher-models

Python
362
star
7

bigdetection

BigDetection: A Large-scale Benchmark for Improved Object Detector Pre-training
Python
352
star
8

earth-forecasting-transformer

Official implementation of Earthformer
Jupyter Notebook
337
star
9

sccl

Pytorch implementation of Supporting Clustering with Contrastive Learning, NAACL 2021
Python
262
star
10

prompt-pretraining

Official implementation for the paper "Prompt Pre-Training with Over Twenty-Thousand Classes for Open-Vocabulary Visual Recognition"
Python
250
star
11

RefChecker

RefChecker provides automatic checking pipeline and benchmark dataset for detecting fine-grained hallucinations generated by Large Language Models.
Python
235
star
12

esci-data

Shopping Queries Dataset: A Large-Scale ESCI Benchmark for Improving Product Search
Python
154
star
13

video-contrastive-learning

Video Contrastive Learning with Global Context, ICCVW 2021
Python
146
star
14

tgl

Python
143
star
15

gan-control

This package provides a pythorch implementation of "GAN-Control: Explicitly Controllable GANs", ICCV 2021.
Jupyter Notebook
122
star
16

polygon-transformer

Python
120
star
17

ReFinED

ReFinED is an efficient and accurate entity linking (EL) system.
Python
116
star
18

unconditional-time-series-diffusion

Official PyTorch implementation of TSDiff models presented in the NeurIPS 2023 paper "Predict, Refine, Synthesize: Self-Guiding Diffusion Models for Probabilistic Time Series Forecasting"
Python
112
star
19

crossnorm-selfnorm

CrossNorm and SelfNorm for Generalization under Distribution Shifts, ICCV 2021
Python
111
star
20

cceval

CrossCodeEval: A Diverse and Multilingual Benchmark for Cross-File Code Completion (NeurIPS 2023)
Python
109
star
21

wqa_tanda

This repo provides code and data used in our TANDA paper.
106
star
22

spot-diff

Project for <SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation> (ECCV 2022)
Python
101
star
23

mintaka

Dataset from the paper "Mintaka: A Complex, Natural, and Multilingual Dataset for End-to-End Question Answering" (COLING 2022)
Python
101
star
24

mix-generation

MixGen: A New Multi-Modal Data Augmentation
Python
100
star
25

long-short-term-transformer

[NeurIPS 2021 Spotlight] Official implementation of Long Short-Term Transformer for Online Action Detection
Python
100
star
26

alexa-arena

Python
99
star
27

fraud-dataset-benchmark

Repository for Fraud Dataset Benchmark
Jupyter Notebook
96
star
28

glass-text-spotting

Official implementation for "GLASS: Global to Local Attention for Scene-Text Spotting" (ECCV'22)
Python
94
star
29

meta-q-learning

Code for the paper "Meta-Q-Learning"( ICLR 2020)
Python
92
star
30

exponential-moving-average-normalization

PyTorch implementation of EMAN for self-supervised and semi-supervised learning: https://arxiv.org/abs/2101.08482
Python
91
star
31

co-with-gnns-example

HTML
88
star
32

datatuner

Code related to "Have Your Text and Use It Too! End-to-End Neural Data-to-Text Generation with Semantic Fidelity" paper
Python
87
star
33

mxeval

Python
84
star
34

sentence-representations

Python
77
star
35

CodeSage

CodeSage: Code Representation Learning At Scale (ICLR 2024)
Python
75
star
36

semimtr-text-recognition

Multimodal Semi-Supervised Learning for Text Recognition (SemiMTR)
Python
75
star
37

fact-check-summarization

Python
72
star
38

instruct-video-to-video

Python
69
star
39

tabsyn

Official Implementations of "Mixed-Type Tabular Data Synthesis with Score-based Diffusion in Latent Space""
Python
68
star
40

object-centric-learning-framework

Python
67
star
41

omni-detr

PyTorch implementation of Omni-DETR for omni-supervised object detection: https://arxiv.org/abs/2203.16089
Python
64
star
42

progressive-coordinate-transforms

Progressive Coordinate Transforms for Monocular 3D Object Detection, NeurIPS 2021
Python
63
star
43

FeatGraph

Python
62
star
44

small-baseline-camera-tracking

A dataset to facilitate the research of Structure-from-Motion (SfM) for movie and TV shows.
61
star
45

tubelet-transformer

This is an official implementation of TubeR: Tubelet Transformer for Video Action Detection
Python
59
star
46

embert

Code for EmBERT, a transformer model for embodied, language-guided visual task completion.
Python
52
star
47

RAGChecker

RAGChecker: A Fine-grained Framework For Diagnosing RAG
Python
52
star
48

probconserv

Datasets and code for results presented in the ProbConserv paper
Python
50
star
49

semi-vit

PyTorch implementation of Semi-supervised Vision Transformers
Python
48
star
50

qa-dataset-converter

Code from the paper "What do Models Learn from Question Answering Datasets?" (EMNLP 2020)
Python
48
star
51

masked-diffusion-lm

Official implementation for the paper "A Cheaper and Better Diffusion Language Model with Soft-Masked Noise"
Python
48
star
52

transformer-gan

Python
47
star
53

transformers-data-augmentation

Code associated with the "Data Augmentation using Pre-trained Transformer Models" paper
Python
46
star
54

gluonmm

A library of transformer models for computer vision and multi-modality research
Python
46
star
55

crossmodal-contrastive-learning

CrossCLR: Cross-modal Contrastive Learning For Multi-modal Video Representations, ICCV 2021
Python
45
star
56

recode

Releasing code for "ReCode: Robustness Evaluation of Code Generation Models"
Python
44
star
57

tracking-dataset

Python
44
star
58

dstc11-track2-intent-induction

DSTC 11 Track 2: Intent Induction from Conversations for Task-Oriented Dialogue
Python
43
star
59

dse

Python
43
star
60

dq-bart

DQ-BART: Efficient Sequence-to-Sequence Model via Joint Distillation and Quantization (ACL 2022)
Python
43
star
61

gnn-tail-generalization

Python
43
star
62

auto-rag-eval

Code repo for the ICML 2024 paper "Automated Evaluation of Retrieval-Augmented Language Models with Task-Specific Exam Generation"
Python
42
star
63

boon

Datasets and code for results presented in the BOON paper
Jupyter Notebook
41
star
64

proteno

This repository contains data used in the NAACL 2021 Paper - Proteno: Text Normalization with Limited Data for Fast Deployment in Text to Speech Systems (https://arxiv.org/abs/2104.07777)
40
star
65

fact-graph

Implementation of the paper "FactGraph: Evaluating Factuality in Summarization with Semantic Graph Representations (NAACL 2022)"
Python
39
star
66

c2f-seg

Official Implementation for ICCV'23 paper Coarse-to-Fine Amodal Segmentation with Shape Prior (C2F-Seg).
Python
38
star
67

amazon-multilingual-counterfactual-dataset

37
star
68

QA-ViT

Python
37
star
69

indoor-scene-generation-eai

Jupyter Notebook
36
star
70

long-tailed-ood-detection

Official implementation for "Partial and Asymmetric Contrastive Learning for Out-of-Distribution Detection in Long-Tailed Recognition" (ICML'22 Long Presentation)
Python
36
star
71

efficient-longdoc-classification

Python
35
star
72

object-centric-multiple-object-tracking

Python
34
star
73

hyperbolic-embeddings

Code for hyperboloid embeddings for knowledge graph entities
Python
33
star
74

domain-knowledge-injection

Python
33
star
75

azcausal

Causal Inference in Python
Python
32
star
76

Repoformer

Repoformer: Selective Retrieval for Repository-Level Code Completion (ICML 2024)
Python
32
star
77

ContraCLM

[ACL 2023] Code for ContraCLM: Contrastive Learning For Causal Language Model
Python
31
star
78

unified-ept

A Unified Efficient Pyramid Transformer for Semantic Segmentation, ICCVW 2021
Python
29
star
79

robust-tableqa

Two approaches for robust TableQA: 1) ITR is a general-purpose retrieval-based approach for handling long tables in TableQA transformer models. 2) LI-RAGE is a robust framework for open-domain TableQA which addresses several limitations. (ACL 2023)
Python
29
star
80

bold

Dataset associated with "BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generation" paper
27
star
81

replay-based-recurrent-rl

Code for "Task-Agnostic Continual RL: In Praise of a Simple Baseline"
Python
26
star
82

controlling-llm-memorization

Python
25
star
83

carbon-assessment-with-ml

CaML: Carbon Footprinting of Household Products with Zero-Shot Semantic Text Similarity
Jupyter Notebook
25
star
84

peft-design-spaces

Official implementation for "Parameter-Efficient Fine-Tuning Design Spaces"
Python
24
star
85

llm-interpret

Code for the ACL 2023 paper: "Rethinking the Role of Scale for In-Context Learning: An Interpretability-based Case Study at 66 Billion Scale"
Python
24
star
86

creating-and-correcting-novel-ml-model-errors

Jupyter Notebook
24
star
87

BartGraphSumm

Implementation of the paper "Efficiently Summarizing Text and Graph Encodings of Multi-Document Clusters (NAACL 2021)"
Python
23
star
88

tofueval

23
star
89

wqa-cascade-transformers

21
star
90

textadain-robust-recognition

TextAdaIN: Paying Attention to Shortcut Learning in Text Recognizers
Python
21
star
91

multiatis

Data and code for the paper "End-to-End Slot Alignment and Recognition for Cross-Lingual NLU" (Accepted to EMNLP 2020)
Python
20
star
92

iwslt-autodub-task

Python
20
star
93

street-reasoning

STREET: a multi-task and multi-step reasoning dataset
Python
19
star
94

contrastive-controlled-mt

Code and data for the IWSLT 2022 shared task on Formality Control for SLT
Ruby
19
star
95

pizza-semantic-parsing-dataset

The PIZZA dataset continues the exploration of task-oriented parsing by introducing a new dataset for parsing pizza and drink orders, whose semantics cannot be captured by flat slots and intents.
Python
19
star
96

redset

Redset is a dataset containing three months worth of user query metadata that ran on a selected sample of instances in the Amazon Redshift fleet. We provide query metadata for 200 provisioned and serverless instances each.
19
star
97

fast-rl-with-slow-updates

Jupyter Notebook
18
star
98

few-shot-baseline

Python
17
star
99

doc-mt-metrics

Python
17
star
100

normalizer-free-robust-training

Official implementation of "Removing Batch Normalization Boosts Adversarial Training" (ICML'22)
Python
17
star