Skip to content
An implementation of the standard AlexNet convolutional network in Swift for TensorFlow
Branch: master
Clone or download
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
AlexNet-Swift.xcodeproj Updated for new loss function that moves away from one-hot labels. Mar 4, 2019
AlexNet Updated for new loss function that moves away from one-hot labels. Mar 4, 2019
images Inserted code for LRN, added an initial toy dataset from Pexels to te… Feb 22, 2019
weights
.gitignore
Package.swift
README.md Updated for new loss function that moves away from one-hot labels. Mar 4, 2019

README.md

AlexNet in Swift

This is an implementation of the classic AlexNet convolutional neural network architecture in Swift for TensorFlow.

It currently builds and runs on both the Mac (via an Xcode project) and on Ubuntu 18.04 (via the swift build command in the main directory). The application has been updated to be current with the Swift for TensorFlow API as of the 0.2 release (March 2, 2019).

On the Mac, you may need to open up your Xcode's File | Project Settings... menu option and switch the Build System to Legacy Build System to get this to compile.

On Ubuntu, running swift build will deposit the binary in the .build directory, so to run this against the weights and images that ship with the project, you'll want to run the following from the base project directory:

.build/x86_64-unknown-linux/debug/AlexNet
You can’t perform that action at this time.