# Neural Machine Translation (NMT)
## Description
This is the Neural Machine Translation package for NLPIA 2nd Edition. Currently support Spanish-English Seq2Seq model using 1-layer GRU with Bag-of-Word accuracy
## Installation
If you just want to install nmt package from PyPI channel:
`console $ pip install nmt==0.0.4 `
If you want to modify the source code to run experiments you’ll need to install dependencis in an environment and then install the package in –editable mode.
### Environment
Dependencies:
NLTK
editdistance
Create a conda environment where you can install all the dependencies like pytorch, pandas, nltk, spacy, and scikit-learn. Jupyter is also installed so developers can experiment in jupyter console (ipython) and Data Scientists can use jupyter notebook.
`console $ conda update -y -n base -c defaults conda $ conda create -y -n nmt 'python=3.7.9' $ conda env update -n nmt -f environment.yml $ conda activate nmt || source activate nmt `
## Usage
### Train an NMT model
Activate conda env with the nmt package installed
nmt –config ${model_hyperparameter_json} –epochs ${num_epoch} –data_path ${training_file} –model_checkpoint_dir ${export_path} –metrics_dir ${metrics_path}
### Parameters
Model Hyperparameter Json: Name of the config file (under the experiment subdirectory)
Epoch: Number of Epoch
Training Text File: Directory of the training corpus (.txt)
Model Checkpoint Path: Directory to save model checkpoint
Metric Directory: Directory to save learning curve and model metrics
## Roadmap - [ ] 0. [Add badge for unittests](https://docs.gitlab.com/ee/user/project/badges.html) to README.md - [ ] 0. Push release to pypi: git tag -a 0.0.6 -m ‘toy_problem.py works!’ && python setup.py sdist bdist_wheel upload - [x] 1. Set up a simple decoder-encoder model using GRU cells, BLEU score as evaluation metrics - [x] 2. Conduct hyperparameter search - [x] 3. Add Attention Mechanism to Decoder-Encoder module - [ ] 4. Incorporate transfer learning from BERT or other models
## Directory structure
Code Structure within source directory: - experiments: submodule where hyperparameters are stored in json format and retrieved as config - models: submodule where Decoder, Encoder, Seq2Seq models are stored - utils: submodule where Word Dictionary and Data Preprocessing functions are found - main_script.py: script to kick start model training - training.py: script to walk through the whole training process
## Credits/References:
[Benjamin Etienne’s repo](https://github.com/b-etienne/Seq2seq-PyTorch/)
[PyTorch’s documentation on Seq2Seq](https://pytorch.org/tutorials/intermediate/seq2seq_translation_tutorial.html)
Metadata
Release files for nmt 0.0.6
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| nmt-0.0.6.tar.gz | 15.7 MB | Without provenance | Details for nmt-0.0.6.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| nmt-0.0.6-py2.py3-none-any.whl | Python 2, Python 3 | none | any | Details for nmt-0.0.6-py2.py3-none-any.whl |
Total release size: 15.8 MB
Release files / nmt-0.0.6.tar.gz
| Download URL | nmt-0.0.6.tar.gz |
|---|---|
| Size | 15.7 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
65db964c0afd5b0dbc4baecdce736f53222f63b791958176b5771db78a005096
|
|
BLAKE2b-256 checksum How to use checksums |
e3861714a0e335a6b32717eab401c7dcc50fe2bb2a65b1f4963af72c5c01a608
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
Python-urllib/3.7
|
Release files / nmt-0.0.6-py2.py3-none-any.whl
| Download URL | nmt-0.0.6-py2.py3-none-any.whl |
|---|---|
| Size | 35.4 kB |
| Tags | Python 2 Python 3 |
|
SHA-256 checksum How to use checksums |
a6d5ee7dbff9695244bbdb2a2b9f60bcb3fc0c05659efb968d1ca49a0bcddfee
|
|
BLAKE2b-256 checksum How to use checksums |
0bb9903db2223856e659edbe3fc86f9c509b0ba159013a2b3f07bf8a36d42114
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
Python-urllib/3.7
|
Security for nmt 0.0.6
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.