Reader-Translator-Generator (RTG)

Reader-Translator-Generator (RTG) is a Neural Machine Translation toolkit based on pytorch. Refer to https://isi-nlp.github.io/rtg/ for the docs.

Features

• Reproducible experiments: one conf.yml that has everything -- data paths, params, and hyper params -- required to reproduce experiments.
• Pre-processing options: sentencepiece or nlcodec (or add your own)
  • word/char/bpe etc types
  • shared vocabulary, seperate vocabulary
  • one-way, two-way, three-way tied embeddings
• Transformer model from "Attention is all you need" (fully tested and competes with Tensor2Tensor
  • Automatically detects and parallelizes across multi GPUs. (Note: All GPUs must be in the same node, though!)
  • Lot of varieties of transformer: width varying, skip transformer etc
• RNN based Encoder-Decoder with Attention . (No longer use it, but it's there for experimentation)
• Language Modeling: RNN, Transformer
• And more ..
  • Easy and interpretable code (for those who read code as much as papers)
  • Object Orientated Design. (Not too many levels of functions and function factories like Tensor2Tensor)
  • Experiments and reproducibility are main focus. To control an experiment you edit an YAML file that is inside the experiment directory.
  • Where ever possible, prefer convention-over-configuation. Have a look at this experiment directory for the examples/transformer.test.yml;

Setup

Add the root of this repo to PYTHONPATH or install it via pip --editable

git clone https://github.com/isi-nlp/rtg-xt.git # use rtg.git if you have access
cd rtg                # go to the code


conda create -n rtg python=3.7   # adds a conda env named rtg
conda activate rtg  # activate it

# install this as a local editable pip package
pip install --editable .   
# All requirements are in setup.py

Usage

Refer to scripts/rtg-pipeline.sh bash script and examples/transformer.base.yml file for specific examples.

The pipeline takes source (.src) and target (.tgt) files. The sources are in one language and the targets in another. At a minimum, supply a training source, training target, validation source, and validation target. It is best to use .tok files for training. (.tok means tokenized.)

Example of training and running a mdoel:

# disable gpu use (force cpu)
export CUDA_VISIBLE_DEVICES=
# call as python module
rtg-pipe experiments/sample-exp/

# OR, you can call a shell scrupt to submit job to slurm/SGE
scripts/rtg-pipeline.sh -d experiments/sample-exp/ -c experiments/sample-exp/conf.yml
# Note: use examples/transformer.base.yml config to setup transformer base

# Then to use the model to translate something:
# (VERY poor translation due to small training data)
echo "Chacun voit midi à sa porte." | rtg-decode experiments/sample-exp/

The 001-tfm directory that hosts an experiment looks like this:

001-tfm
├── _PREPARED    <-- Flag file indicating experiment is prepared 
├── _TRAINED     <-- Flag file indicating experiment is trained
├── conf.yml     <-- Where all the params and hyper params are! You should look into this
├── data        
│   ├── samples.tsv.gz          <-- samples to log after each check point during training
│   ├── sentpiece.shared.model  <-- as the name says, sentence piece model, shared
│   ├── sentpiece.shared.vocab  <-- as the name says
│   ├── train.db                <-- all the prepared trainig data in a sqlite db
│   └── valid.tsv.gz            <-- and the validation data
├── githead       <-- whats was the git HEAD hash this experiment was started? 
├── job.sh.bak    <-- job script used to submit this to grid. Just in case
├── models        <-- All checkpoints go inside this
│   ├── model_400_5.265583_4.977106.pkl
│   ├── model_800_4.478784_4.606745.pkl
│   ├── ...
│   └── scores.tsv <-- train and validation losses. incase you dont want to see tensorboard
├── rtg.log   <-- the python logs are redirected here
├── rtg.zip   <-- the source code used to run. just `export PYTHONPATH=rtg.zip` to 
├── scripts -> /Users/tg/work/me/rtg/scripts  <-- link to some perl scripts for detok+BLEU
├── tensorboard    <-- Tensorboard stuff for visualizations
│   ├── events.out.tfevents.1552850552.hackb0x2
│   └── ....
└── test_step2000_beam4_ens5   <-- Tests after the end of training, BLEU scores
    ├── valid.ref -> /Users/tg/work/me/rtg/data/valid.ref
    ├── valid.src -> /Users/tg/work/me/rtg/data/valid.src
    ├── valid.out.tsv
    ├── valid.out.tsv.detok.tc.bleu
    └── valid.out.tsv.detok.lc.bleu

Authors:

See Here

Credits / Thanks

• OpenNMT and the Harvard NLP team for Annotated transformer, I learned a lot from their work
• My team at USC ISI for everything else