Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/jackroos/VL-BERT

Code for ICLR 2020 paper "VL-BERT: Pre-training of Generic Visual-Linguistic Representations".
https://github.com/jackroos/VL-BERT

bert iclr2020 pre-training pytorch representation-learning self-supervised-learning vision-and-language vl-bert

Last synced: 6 days ago
JSON representation

Code for ICLR 2020 paper "VL-BERT: Pre-training of Generic Visual-Linguistic Representations".

Host: GitHub
URL: https://github.com/jackroos/VL-BERT
Owner: jackroos
License: mit
Created: 2019-11-22T06:08:58.000Z (almost 5 years ago)
Default Branch: master
Last Pushed: 2023-05-22T22:33:35.000Z (over 1 year ago)
Last Synced: 2024-07-21T12:35:03.064Z (4 months ago)
Topics: bert, iclr2020, pre-training, pytorch, representation-learning, self-supervised-learning, vision-and-language, vl-bert
Language: Jupyter Notebook
Homepage:
Size: 5.41 MB
Stars: 735
Watchers: 14
Forks: 111
Open Issues: 20
Metadata Files:
- Readme: README.md
- License: LICENSE

Awesome Lists containing this project

awesome-Vision-and-Language-Pre-training - [code
awesome-vision-language-pretraining-papers - [code
StarryDivineSky - jackroos/VL-BERT

README

# VL-BERT

By
[Weijie Su](https://www.weijiesu.com/),
[Xizhou Zhu](https://scholar.google.com/citations?user=02RXI00AAAAJ&hl=en),
[Yue Cao](http://yue-cao.me/),
[Bin Li](http://staff.ustc.edu.cn/~binli/),
[Lewei Lu](https://www.linkedin.com/in/lewei-lu-94015977/),
[Furu Wei](http://mindio.org/),
[Jifeng Dai](https://jifengdai.org/).

This repository is an official implementation of the paper
[VL-BERT: Pre-training of Generic Visual-Linguistic Representations](https://arxiv.org/abs/1908.08530).

*Update on 2020/01/16* Add code of visualization.

*Update on 2019/12/20* Our VL-BERT got accepted by ICLR 2020.

## Introduction

VL-BERT is a simple yet powerful pre-trainable generic representation for visual-linguistic tasks.
It is pre-trained on the massive-scale caption dataset and text-only corpus,
and can be fine-tuned for various down-stream visual-linguistic tasks,
such as Visual Commonsense Reasoning, Visual Question Answering and Referring Expression Comprehension.

![](./figs/pretrain.png)

![](./figs/attention_viz.png)

Thanks to PyTorch and its 3rd-party libraries, this codebase also contains following features:
* Distributed Training
* FP16 Mixed-Precision Training
* Various Optimizers and Learning Rate Schedulers
* Gradient Accumulation
* Monitoring the Training Using TensorboardX

## Citing VL-BERT
```bibtex
@inproceedings{
Su2020VL-BERT:,
title={VL-BERT: Pre-training of Generic Visual-Linguistic Representations},
author={Weijie Su and Xizhou Zhu and Yue Cao and Bin Li and Lewei Lu and Furu Wei and Jifeng Dai},
booktitle={International Conference on Learning Representations},
year={2020},
url={https://openreview.net/forum?id=SygXPaEYvH}
}
```

## Prepare

### Environment
* Ubuntu 16.04, CUDA 9.0, GCC 4.9.4
* Python 3.6.x
```bash
# We recommend you to use Anaconda/Miniconda to create a conda environment
conda create -n vl-bert python=3.6 pip
conda activate vl-bert
```
* PyTorch 1.0.0 or 1.1.0
```bash
conda install pytorch=1.1.0 cudatoolkit=9.0 -c pytorch
```
* Apex (optional, for speed-up and fp16 training)
```bash
git clone https://github.com/jackroos/apex
cd ./apex
pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./
```
* Other requirements:
```bash
pip install Cython
pip install -r requirements.txt
```
* Compile
```bash
./scripts/init.sh
```

### Data

See [PREPARE_DATA.md](data/PREPARE_DATA.md).

### Pre-trained Models

See [PREPARE_PRETRAINED_MODELS.md](model/pretrained_model/PREPARE_PRETRAINED_MODELS.md).

## Training

### Distributed Training on Single-Machine

```
./scripts/dist_run_single.sh /train_end2end.py
```
* ``````: number of gpus to use.
* ``````: pretrain/vcr/vqa/refcoco.
* ``````: config yaml file under ```./cfgs/```.
* ``````: root directory to store checkpoints.

Following is a more concrete example:
```
./scripts/dist_run_single.sh 4 vcr/train_end2end.py ./cfgs/vcr/base_q2a_4x16G_fp32.yaml ./
```

### Distributed Training on Multi-Machine

For example, on 2 machines (A and B), each with 4 GPUs,

run following command on machine A:
```
./scripts/dist_run_multi.sh 2 0 4 /train_end2end.py
```

run following command on machine B:
```
./scripts/dist_run_multi.sh 2 1 4 /train_end2end.py
```

### Non-Distributed Training
```
./scripts/nondist_run.sh /train_end2end.py
```

***Note***:

1. In yaml files under ```./cfgs```, we set batch size for GPUs with at least 16G memory, you may need to adapt the batch size and
gradient accumulation steps according to your actual case, e.g., if you decrease the batch size, you should also
increase the gradient accumulation steps accordingly to keep 'actual' batch size for SGD unchanged.

2. For efficiency, we recommend you to use distributed training even on single-machine. But for RefCOCO+, you may meet deadlock
using distributed training due to unknown reason (it may be related to [PyTorch dataloader deadloack](https://github.com/pytorch/pytorch/issues/1355)), you can simply use
non-distributed training to solve this problem.

## Evaluation

### VCR
* Local evaluation on val set:
```
python vcr/val.py \
--a-cfg --r-cfg \
--a-ckpt --r-ckpt \
--gpus \
--result-path --result-name
```
***Note***: `````` is gpu indexes, e.g., ```0 1 2 3```.

* Generate prediction results on test set for [leaderboard submission](https://visualcommonsense.com/leaderboard/):
```
python vcr/test.py \
--a-cfg --r-cfg \
--a-ckpt --r-ckpt \
--gpus \
--result-path --result-name
```

### VQA
* Generate prediction results on test set for [EvalAI submission](https://evalai.cloudcv.org/web/challenges/challenge-page/163/overview):
```
python vqa/test.py \
--cfg \
--ckpt \
--gpus \
--result-path --result-name
```

### RefCOCO+

* Local evaluation on val/testA/testB set:
```
python refcoco/test.py \
--split \
--cfg \
--ckpt \
--gpus \
--result-path --result-name
```

## Visualization
See [VISUALIZATION.md](./viz/VISUALIZATION.md).

## Acknowledgements

Many thanks to following codes that help us a lot in building this codebase:
* [transformers (pytorch-pretrained-bert)](https://github.com/huggingface/transformers)
* [Deformable-ConvNets](https://github.com/msracver/Deformable-ConvNets/)
* [maskrcnn-benchmark](https://github.com/facebookresearch/maskrcnn-benchmark)
* [mmdetection](https://github.com/open-mmlab/mmdetection)
* [r2c](https://github.com/rowanz/r2c)
* [allennlp](https://github.com/allenai/allennlp)
* [bottom-up-attention](https://github.com/peteanderson80/bottom-up-attention)
* [pythia](https://github.com/facebookresearch/pythia)
* [MAttNet](https://github.com/lichengunc/MAttNet)
* [bertviz](https://github.com/jessevig/bertviz)