Metadata-Version: 2.1
Name: reformer-fastai
Version: 0.0.2
Summary: Fastai community entry to 2020 Reproducibility Challenge
Home-page: https://github.com/arampacha/reformer_fastai/tree/master/
Author: Fastai community
Author-email: aruthart@gmail.com
License: MIT License
Keywords: fastai reformer
Platform: UNKNOWN
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Natural Language :: English
Classifier: Programming Language :: Python :: 3.6
Classifier: Programming Language :: Python :: 3.7
Classifier: Programming Language :: Python :: 3.8
Requires-Python: >=3.6
Description-Content-Type: text/markdown
Requires-Dist: torch (>=1.7.0)
Requires-Dist: fastai (<2.3,>=2.2.2)
Requires-Dist: fastcore (>=1.3.19)
Requires-Dist: einops
Requires-Dist: torchvision (>=0.8)
Requires-Dist: spacy (>=2.2.4)
Requires-Dist: matplotlib
Requires-Dist: axial-positional-embedding
Requires-Dist: datasets

# Reformer Reproducibility Experiments
> Fastai community entry to <a href='https://paperswithcode.com/rc2020'>2020 Reproducibility Challenge</a>


## Installation

### Setup

If you don't already, its a good idea to install the package into a virtual environment

```
python3 -m venv my_env

source ./my_env/bin/activate
```

### Run Install
Then you can install the package by running:

`pip install git+git://github.com/arampacha/reformer_fastai.git`

## Project Links

- [Older repo](https://github.com/morganmcg1/reformer-fastai-old)
- [Repo docs](https://arampacha.github.io/reformer_fastai/)
- [Fastai forums thread](https://forums.fast.ai/t/reproducibility-challenge-2020-fastai-folks-interested/80336/39)
- [Google doc](https://docs.google.com/document/d/1wF83E3B3yXIGZixEgOUJI2T2XXhT1DVCrPXS5Dbsyh8/edit)
- [RC2020 Report Style Template](https://paperswithcode.com/static/rc2020/ML-Reproducibility-Challenge-2020-Template.zip)

## Resources

### Author's Code and Resources
- [Reformer Paper](https://openreview.net/pdf?id=rkgNKkHtvB)
- [Authors ICLR video](https://iclr.cc/virtual_2020/poster_rkgNKkHtvB.html)
- [Google Blog](https://ai.googleblog.com/2020/01/reformer-efficient-transformer.html)
- [Authors code (TRAX)](https://github.com/google/trax/tree/master/trax/models/reformer)
- [Reformer enwik8 model and training config](https://github.com/google/trax/blob/f8024e8057599b92fce82842f342cb3d39c8f405/trax/supervised/configs/reformer_enwik8.gin)

### More Code
- [@lucidrain’s Reformer code](https://github.com/lucidrains/reformer-pytorch/)
- [HuggingFace: Reformer source code](https://github.com/huggingface/transformers/blob/a1bbcf3f6c20e15fe799a8659d6b7bd36fdf11ed/src/transformers/modeling_reformer.py)
    - [Reformer enwik8 configs](https://github.com/google/trax/blob/master/trax/supervised/configs/reformer_enwik8.gin)
    - [Reformer WMT14 en-de configs](https://github.com/google/trax/blob/master/trax/supervised/configs/reformer_wmt_ende.gin)
    - [Reformer Machine Translation example](https://github.com/google/trax/blob/a0483a12cb7ebece40b5e302e8e81fd9249c6ef6/trax/models/reformer/machine_translation.ipynb)
    = [SubwordTextEncoder tokenizer used for Machine Translation](https://github.com/tensorflow/tensor2tensor/blob/21dba2c1bdcc7ab582a2bfd8c0885c217963bb4f/tensor2tensor/data_generators/text_encoder.py#L448)
- [HuggingFace: Reformer notebook example](https://colab.research.google.com/github/patrickvonplaten/blog/blob/master/notebooks/03_reformer.ipynb)
- [HuggingFace: long sequences](https://colab.research.google.com/github/patrickvonplaten/notebooks/blob/master/PyTorch_Reformer.ipynb)
- [HuggingFace: Pretraining](https://colab.research.google.com/drive/1tzzh0i8PgDQGV3SMFUGxM7_gGae3K-uW?usp=sharing)

### Data

**enwik8**
- [enwik8.zip, raw data, 100mb](http://mattmahoney.net/dc/enwik8.zip)
- [Tensor2Tensor enwik8 data generator code, with train/dev/test split](https://github.com/tensorflow/tensor2tensor/blob/master/tensor2tensor/data_generators/enwik8.py). File lengths:
    - Train: 89,621,832
    - Eval: 5,000,000
    - Test: 5,000,000
- [enwik8 notebook Tensor2Tensor](https://github.com/morganmcg1/reformer-fastai/blob/main/enwiki8_Tensor2Tensor_download.ipynb)

**WMT14**
- [WMT on HuggingFace Datasets](https://huggingface.co/datasets/viewer/?dataset=wmt14&config=cs-en)
- [Reformer WMT14 vocab](https://github.com/google/trax/tree/a0483a12cb7ebece40b5e302e8e81fd9249c6ef6/trax/models/reformer/testdata)
- Reformer.input_vocab_size = 33300, from [WMT14 model config](https://github.com/google/trax/blob/master/trax/supervised/configs/reformer_wmt_ende.gin)
- Train Test split: (guess) newstest2013 for validation and newstest2014 for test, in consistence with Vaswani et al. (2017) - from https://arxiv.org/pdf/2009.02070.pdf
- Tokenizer: [Tensor2Tensor SubWordTextEncoder](https://github.com/tensorflow/tensor2tensor/blob/21dba2c1bdcc7ab582a2bfd8c0885c217963bb4f/tensor2tensor/data_generators/text_encoder.py#L448)

### Explainers

- [Yannic K explainer](https://www.youtube.com/watch?v=i4H0kjxrias&t=1s)
- [HuggingFace blog post](https://huggingface.co/blog/reformer)
- [Illustrating the Reformer blog post](https://towardsdatascience.com/illustrating-the-reformer-393575ac6ba0)
- [Reformer Deep Dive](https://www.pragmatic.ml/reformer-deep-dive/)

### Related

- [Coursera Attention Models in NLP course, with Reformer co-author](https://www.coursera.org/learn/attention-models-in-nlp)
- [@hallvagi Attention blogpost](https://hallvagi.github.io/dl-explorer/fastai/attention/lstm/2020/06/29/Attention.html)
- [The Transformer Family by @lilianweng](https://lilianweng.github.io/lil-log/2020/04/07/the-transformer-family.html)
- [A Family of Attention Mechanisms bu @lilianweng](https://lilianweng.github.io/lil-log/2018/06/24/attention-attention.html#a-family-of-attention-mechanisms)


