# hollance/Forge

A neural network toolkit for Metal

Repository: https://github.com/hollance/Forge
Canonical: https://ross.abutalabs.com/products/hollance-forge
Language: Swift
License: MIT
License Family: permissive
Topics: metal, deep-learning, deep-neural-networks, neural-network, ios, swift, mobilenets, machine-learning
Last push: 2018-05-18T20:25:35+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": 3419, "days_push": 3029, "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 1264, forks 171 (observed 2026-08-28T04:04:10.521020+00:00)

## What it is
Forge is a Swift library that wraps Apple's MPSCNN (Metal Performance Shaders CNN) framework to make it easier to build and run deep neural networks on iOS and macOS GPUs. It provides a DSL for layer creation, data conversion helpers, custom layers like depth-wise convolution, and pretrained example models such as LeNet-5, Inception, and MobileNets.

## Use cases
- build a neural network on iOS using Metal performance shaders
- run MobileNet or SSD inference fast on iPhone GPU
- convert MPSImage and MTLTexture data to Swift Float arrays
- define CNN layers with a simple Swift DSL instead of MPSCNN boilerplate
- implement custom Metal compute kernels for deep learning layers
- debug and inspect intermediate outputs of a Metal neural network

## When to choose
- you need GPU-accelerated neural network inference on iOS 10/11 faster than Core ML
- you want to learn how CNNs are implemented with Metal performance shaders
- you need custom layers like depth-wise convolution on top of MPSCNN

## When to avoid
- you target modern iOS versions, since the MPSCNN API it uses is no longer supported
- you want a maintained library, as the author explicitly no longer maintains it
- Core ML meets your performance needs on current Apple platforms

## Facets
- artifact type: library
- maturity: abandoned
- function: deep-learning, machine-learning, image-processing, developer-tools
- domain: deep-learning, machine-learning, mobile-development, apple-ecosystem
- platform: -
- tags: metal, mps-cnn, neural-networks, core-ml, mobilenet, gpu-acceleration, unmaintained, ios, macos, swift, gpu

## Member repositories
- hollance/Forge (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:10.521020+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 2026-08-30T05:04:03.022102+00:00, confidence not recorded.
  - readme: https://github.com/hollance/Forge (fetched 2026-08-28T04:04:10.521020+00:00, sha 9ef51216a08c)
- Data as of 2026-08-30T08:39:29.467469+00:00.