khakhulin/compressed-transformer
Compression of NMT transformer model with tensor methods
This project helps machine translation practitioners create smaller, more efficient neural machine translation (NMT) models, specifically those built with the Transformer architecture. It takes existing NMT models and applies tensor compression techniques to significantly reduce their size. The output is a functionally equivalent NMT model that requires fewer computational resources for training and deployment.
No commits in the last 6 months.
Use this if you are developing or deploying neural machine translation systems and need to reduce the model size, improve inference speed, or manage memory constraints without sacrificing translation quality.
Not ideal if you are working with non-Transformer based NLP models or if your primary goal is to achieve the absolute highest translation accuracy, as some compression methods may slightly impact performance.
Stars
48
Forks
9
Language
Jupyter Notebook
License
MIT
Category
Last pushed
Jun 07, 2019
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/nlp/khakhulin/compressed-transformer"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
xv44586/toolkit4nlp
transformers implement (architecture, task example, serving and more)
luozhouyang/transformers-keras
Transformer-based models implemented in tensorflow 2.x(using keras).
ufal/neuralmonkey
An open-source tool for sequence learning in NLP built on TensorFlow.
graykode/xlnet-Pytorch
Simple XLNet implementation with Pytorch Wrapper
uzaymacar/attention-mechanisms
Implementations for a family of attention mechanisms, suitable for all kinds of natural language...