# meta-pytorch/tnt

A lightweight library for PyTorch training tools and utilities

Repository: https://github.com/meta-pytorch/tnt
Canonical: https://ross.abutalabs.com/products/tnt
Homepage: https://pytorch.org/tnt/
Language: Python
License: NOASSERTION
License Family: other
Topics: pytorch, neural-network, python, machine-learning, deep-learning, reinforcement-learning
Last push: 2026-08-18T00:06:41+00:00

## Health v2 (maintenance only)
Score: 76/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 98, release rhythm 35, longevity 100
- inputs: {"age_days": 3553, "days_push": 16, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1721, forks 309 (observed 2026-08-28T04:05:27.714707+00:00)

## What it is
TNT (torchtnt) is a lightweight library from Meta providing tools and utilities for PyTorch training workflows. It offers building blocks for training loops, logging, checkpointing, and other common deep-learning training needs.

## Use cases
- train pytorch models with a reusable training loop
- add logging and checkpointing to pytorch training
- simplify pytorch experiment boilerplate
- run distributed pytorch training jobs
- build custom training pipelines for deep learning

## When to choose
- you use PyTorch and want lightweight training utilities without a heavy framework
- you need composable building blocks for training loops, logging, and checkpointing
- you want a Meta/PyTorch-maintained library with active releases

## When to avoid
- you need a full high-level training framework like PyTorch Lightning with callbacks and trainers out of the box
- you train models in TensorFlow or JAX
- you only need inference or deployment tooling rather than training

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-training
- domain: machine-learning, deep-learning, developer-tools
- platform: python, cross-platform
- tags: pytorch, training-loop, training-utilities, neural-networks, reinforcement-learning, gpu

## Member repositories
- meta-pytorch/tnt (main) score 76

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:27.714707+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 2026-08-30T03:34:02.656006+00:00, confidence not recorded.
  - readme: https://github.com/meta-pytorch/tnt (fetched 2026-08-28T04:05:27.714707+00:00, sha a6c3d3be80e0)
  - homepage: https://pytorch.org/tnt/ (fetched 2026-08-29T11:09:46.531151+00:00, sha 3c67ea9f2eb9)
- Data as of 2026-08-30T08:39:29.467469+00:00.