# soumith/convnet-benchmarks

Easy benchmarking of all publicly accessible implementations of convnets

Repository: https://github.com/soumith/convnet-benchmarks
Canonical: https://ross.abutalabs.com/products/convnet-benchmarks
Language: Python
License: MIT
License Family: permissive
Last push: 2017-06-09T15:12:02+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 4435, "days_push": 3372, "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 2686, forks 571 (observed 2026-08-28T04:07:10.751938+00:00)

## What it is
A benchmarking suite and published results comparing publicly available implementations of convolutional neural networks (convnets) across deep learning frameworks like Torch, Caffe, TensorFlow, and Chainer. It measures forward and backward pass times on popular ImageNet models using NVIDIA GPUs.

## Use cases
- compare speed of deep learning frameworks for convnets
- benchmark CNN convolution implementations on GPU
- find the fastest library for AlexNet forward and backward passes
- evaluate CuDNN vs native convolution kernel performance
- measure convnet training throughput across frameworks

## When to choose
- you need historical benchmark data comparing pre-2017 deep learning frameworks
- you want reference timing methodology for convnet benchmarks
- you are researching performance of older frameworks like Torch, Caffe, or Nervana neon

## When to avoid
- you need up-to-date benchmarks of modern frameworks like PyTorch 2.x or JAX
- you want a maintained benchmarking tool for current hardware
- you need benchmarks beyond convolutional networks

## Facets
- artifact type: learning-resource
- maturity: abandoned
- function: benchmarking, machine-learning, deep-learning
- domain: deep-learning, performance, gpu-computing
- platform: python
- tags: convnets, cnn, benchmark-results, deep-learning-frameworks, nvidia-gpu, linux, gpu

## Member repositories
- soumith/convnet-benchmarks (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:10.751938+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 2026-08-30T02:17:47.946986+00:00, confidence not recorded.
  - readme: https://github.com/soumith/convnet-benchmarks (fetched 2026-08-28T04:07:10.751938+00:00, sha 16d1433d574a)
- Data as of 2026-08-30T08:39:29.467469+00:00.