# NVIDIA/apex

A PyTorch Extension:  Tools for easy mixed precision and distributed training in Pytorch

Repository: https://github.com/NVIDIA/apex
Canonical: https://ross.abutalabs.com/products/apex
Language: Python
License: BSD-3-Clause
License Family: permissive
Last push: 2026-08-24T21:31:00+00:00

## Health v2 (maintenance only)
Score: 77/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 35, longevity 100
- inputs: {"age_days": 3054, "days_push": 9, "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 8993, forks 1525 (observed 2026-08-28T04:10:26.692719+00:00)

## What it is
NVIDIA-maintained PyTorch extension providing utilities for easy mixed precision and distributed training. It offers up-to-date CUDA and C++ extensions that eventually make their way into upstream PyTorch.

## Use cases
- train deep learning models with mixed precision in pytorch
- speed up pytorch training on nvidia gpus
- run distributed training across multiple gpus
- use fused cuda kernels for faster optimizer steps
- reduce gpu memory usage during model training

## When to choose
- you train large models on NVIDIA GPUs and need AMP/fused optimizer performance
- you need cutting-edge training utilities before they land in PyTorch
- you use NVIDIA NGC PyTorch containers and want matching extensions

## When to avoid
- you only train on CPU or non-NVIDIA hardware
- you need a stable, simple setup without compiling CUDA extensions
- your use case is fully covered by native PyTorch AMP

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, llm-training, gpu-computing
- domain: deep-learning, machine-learning, gpu-computing
- platform: python
- tags: pytorch-extension, mixed-precision, distributed-training, cuda-extensions, nvidia, linux, gpu

## Member repositories
- NVIDIA/apex (main) score 77

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:26.692719+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 2026-08-29T17:24:22.506962+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA/apex (fetched 2026-08-28T04:10:26.692719+00:00, sha 2e7aea0520a7)
- Data as of 2026-08-30T08:39:29.467469+00:00.