Skip to content

Example tensorflow graphs built in F#, weights loaded from a pre-existing models

License

Notifications You must be signed in to change notification settings

moloneymb/TF_examples

Repository files navigation

Tensorflow Examples in F#

List of examples

  • ImageClassifier
    • Uses Resnet50 to classifiy an image into one of 1000 categories
  • NeuralStyleTransfer
    • Uses Fast Style Transfer to transfer image styles
    • The pretrained weights are for style of Rain Princess by Lenoid Afremov
    • Different weights can be loaded to apply different styles
  • AttGAN
    • Using a GAN to visually change the attributs of faces
    • This is a work in progress

Setup

  • Make sure FSI is 64bit and does not shadow assemblies1
  • run script setup.fsx

Run

  • Scripts can be run by either "Send to FSI" or by FSI. For example FSI.exe NeuralStyleTransfer.fsx --style wave

1: Making sure FSI is 64bit and does not shadow

For Visual Studio Code:

  • On Ubuntu: it should work with default settings
  • On Windows: if FSI is defaulting to 32bit this will need to be changed to 64bit by setting the Ionide property for fsiFilePath to the full path to the fsiAnyCpu.exe.

"FSharp.fsiFilePath": "C:\..fsiAnyCpu.exe"

See Ionide issue [956](https://github.com/ionide/ionide-vscode-fsharp/issues/956)
  • On Mac: this is currently untested

For Visual Studio both 64bit needs to be enabled and shadow copy assemblies needs to be switched off. Tools -> Options -> F# Tools -> F# Interactive

About

Example tensorflow graphs built in F#, weights loaded from a pre-existing models

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages