# harvardnlp/seq2seq-attn

Sequence-to-sequence model with LSTM encoder/decoders and attention

Repository: https://github.com/harvardnlp/seq2seq-attn
Canonical: https://ross.abutalabs.com/products/seq2seq-attn
Homepage: http://nlp.seas.harvard.edu/code
Language: Lua
License: MIT
License Family: permissive
Last push: 2020-12-30T02:54:09+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3807, "days_push": 2072, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1281, forks 277 (observed 2026-08-28T04:04:13.924493+00:00)

## What it is
A Torch (Lua) implementation of sequence-to-sequence models with LSTM encoder-decoders and optional attention for neural machine translation. It supports bidirectional encoders, character-level inputs via CNN+highway networks, and many NMT research extensions.

## Use cases
- train a neural machine translation model
- build a sequence-to-sequence model with attention
- experiment with character-level NMT
- reproduce Luong attention results
- learn how seq2seq encoder-decoder models work

## When to choose
- you need the original reference implementation of Luong attention for research
- you work with Torch/Lua and want a classic seq2seq baseline
- you want to study or extend NMT techniques like pruning or knowledge distillation

## When to avoid
- you want a maintained, production-ready translation system - use OpenNMT instead
- your stack is PyTorch or TensorFlow
- you need modern transformer-based models

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, nlp, deep-learning
- domain: machine-learning, deep-learning
- platform: lua
- tags: seq2seq, attention, lstm, neural-machine-translation, torch, encoder-decoder, character-level, natural-language-processing, gpu, linux, macos

## Member repositories
- harvardnlp/seq2seq-attn (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:13.924493+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:58:17.785318+00:00, confidence not recorded.
  - readme: https://github.com/harvardnlp/seq2seq-attn (fetched 2026-08-28T04:04:13.924493+00:00, sha a0aac914eab9)
  - homepage: http://nlp.seas.harvard.edu/code (fetched 2026-08-29T12:12:54.541018+00:00, sha 4805315d6206)
- Data as of 2026-08-30T08:39:29.467469+00:00.