# pytorch/benchmark

TorchBench is a collection of open source benchmarks used to evaluate PyTorch performance.

Repository: https://github.com/pytorch/benchmark
Canonical: https://ross.abutalabs.com/products/pytorch-benchmark
Language: Python
License: BSD-3-Clause
License Family: permissive
Topics: pytorch, benchmark
Last push: 2026-09-02T15:14:11+00:00

## Health v2 (maintenance only)
Score: 77/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 100, release rhythm 35, longevity 100
- inputs: {"age_days": 3386, "days_push": 0, "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 1041, forks 342 (observed 2026-09-03T02:15:17.152952+00:00)

## What it is
TorchBench is a collection of open source benchmarks used to evaluate PyTorch performance across popular model workloads. It provides a standardized API for benchmark drivers, supports backends like torchinductor/torchscript, and can be installed as a library or run as a suite.

## Use cases
- benchmark pytorch model training and inference performance
- compare performance across pytorch versions or backends
- evaluate torchinductor or torchscript speedups
- measure gpu vs cpu performance of standard models
- regression-test pytorch nightly performance
- run standardized workloads like densenet121 for profiling

## When to choose
- you need standardized, reproducible PyTorch performance benchmarks
- you are comparing PyTorch builds, backends, or hardware
- you contribute to PyTorch performance work

## When to avoid
- you need a general-purpose ML framework rather than benchmarks
- you only need micro-benchmarks of individual ops (use torch.utils.benchmark)
- you require stable-release-only support (nightlies are the tested target)

## Facets
- artifact type: dataset
- maturity: active
- function: benchmarking, machine-learning, deep-learning
- domain: machine-learning, deep-learning, performance, developer-tools
- platform: python, windows
- tags: pytorch, torchbench, model-benchmarks, performance-evaluation, gpu-benchmarking, linux, macos, gpu

## Member repositories
- pytorch/benchmark (main) score 77

## Provenance
- Observed fields: from GitHub, fetched 2026-09-03T02:15:17.152952+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 2026-08-30T07:02:31.717053+00:00, confidence not recorded.
  - readme: https://github.com/pytorch/benchmark (fetched 2026-09-03T02:15:17.152952+00:00, sha 78dd7ad6ffcb)
- Data as of 2026-08-30T08:39:29.467469+00:00.