# namisan/mt-dnn

Multi-Task Deep Neural Networks for Natural Language Understanding

Repository: https://github.com/namisan/mt-dnn
Canonical: https://ross.abutalabs.com/products/mt-dnn
Language: Python
License: MIT
License Family: permissive
Topics: multi-task-learning, natural-language-understanding, deep-learning, microsoft, ranking, named-entity-recognition, bert, machine-reading-comprehension, nlp, pytorch
Last push: 2024-03-07T16:23:07+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": 2752, "days_push": 909, "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 2257, forks 408 (observed 2026-08-28T04:06:31.454828+00:00)

## What it is
A PyTorch library implementing Multi-Task Deep Neural Networks (MT-DNN) for natural language understanding, based on Microsoft research papers. It supports multi-task fine-tuning of pretrained language models like BERT, with techniques such as knowledge distillation, adversarial training (ALUM), and SMART regularization.

## Use cases
- fine-tune BERT on multiple NLU tasks jointly
- train a model for GLUE benchmark tasks
- do multi-task learning for text classification and NER
- apply adversarial training to language model fine-tuning
- use knowledge distillation to improve NLU models
- implement machine reading comprehension models
- reproduce ACL/NAACL research on robust NLU fine-tuning

## When to choose
- you want multi-task fine-tuning of pretrained transformers in PyTorch
- you need research-grade implementations of ALUM, SMART, or R-Drop style regularization
- you are reproducing the MT-DNN papers or GLUE-style experiments

## When to avoid
- you need pretrained model checkpoints - public model hosting was discontinued
- you want a maintained production NLP toolkit - development is largely inactive
- you need support for modern Python versions - the code targets older Python 3.6 environments

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, nlp, deep-learning
- domain: machine-learning, deep-learning
- platform: python
- tags: multi-task-learning, bert, pytorch, knowledge-distillation, adversarial-training, text-classification, named-entity-recognition, question-answering, natural-language-processing, linux, docker

## Member repositories
- namisan/mt-dnn (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:31.454828+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 2026-08-30T02:43:40.367288+00:00, confidence not recorded.
  - readme: https://github.com/namisan/mt-dnn (fetched 2026-08-28T04:06:31.454828+00:00, sha 9fa11f6ef7fc)
- Data as of 2026-08-30T08:39:29.467469+00:00.