F# samples need to be migrated to ML.NET v0.7 (minor migration from v0.6)

[CESARDELATORRE]

We released v0.7 today (Nov. 6th) and there are a few minor breaking changes compared to v0.6 in some Estimators naming, most of all.
For now, I'm setting a special "Directory.Build.props" file targeting v0.6 for the F# samples until they are migrated to v0.7.

C# samples are most of them already migrated to v0.7, so contributors for F# can check it out in regards naming, etc.

[kunjee17]

@mariuszwojcik let's go for it. Which one you are picking up?

@CESARDELATORRE is it ok if we use paket instead of nuget for F# samples ? F# community like that way. If you want I can open up new issue to discuss this.

[CESARDELATORRE]

Sure, open an issue and let's discuss it with @dsyme ok? 👍

[mariuszwojcik]

@kunjee17 hey, welcome back, hope you had great holiday. The basic four samples are already on v0.7. I am working on porting "Fraud detection" to F#, so if you wish you could port any other sample.

[kunjee17]

@mariuszwojcik is your samples are already merged ? I ll pick it others. And also mention them here so we can stay in sync.

[mariuszwojcik]

@kunjee17 when you get the latest, all F# samples are in v0.7 and as I said, I am working on Fraud Detection

[CESARDELATORRE]

@mariuszwojcik @Kunjee Hey guys! Thanks a lot for your work migrating the F# samples. We really appreciate it from Microsoft! 👍 - About continuing with further samples for F#, just a heads-up.
Honoring your time, It'd be better if you still don't migrate ALL the C# samples, as some of them still will be refactored quite a lot. Especially the ones who are still not using the "common code".

I'd suggest if you want to migrate the following three samples, only, for now:

• Regression_BikeSharingDemand
• MulticlassClassification-GitHubLabeler
• Clustering_CustomerSegmentation

And even those (and the rest) might slightly be refactored in the upcoming days while reviewing the samples from an API point of view.

But do NOT migrate the following as they will evolve quite a lot until we align them to the rest of samples:

• eShopDashboardML
• BinaryClassification_CreditCardFraudDetection
• DeepLearning_ImageClassification_TensorFlow
• DeepLearning_TensorFlowEstimator

Thanks! 👍

[mariuszwojcik]

@CESARDELATORRE thanks for the heads up - much appreciated. In that case I will focus my efforts on Regression_BikeSharingDemand for now.

Cheers!

[mariuszwojcik]

Regression_BikeSharingDemand sample has been migrated to F#. Next one to go would be MulticlassClassification-GitHubLabeler

Cheers

[kunjee17]

I ll try for Clustering_CustomerSegmentation.

[kunjee17]

@mariuszwojcik it would be good to separate common folder that can be reused ? I am as of now following your work and using your code only. But may be in future a nuget helper package would be good ?

[mariuszwojcik]

@kunjee17 there already is a common folder which is reused between all samples. You'll find there ModelBuilder and ModelScorer modules as well as ConsoleHelper. This follows C# samples and I think keeping them in similar structure will be beneficial for people learning how to use ML.NET as they will be able to easily follow C# and F# samples.
I was thinking to move several samples to F#, see what code requires fsharp-isation and then move it to dedicated assembly which can be reused. The first attempt was a Pipeline module, but since then the logic moved to the ModelBuilder. Currently, the project changes a lot so it may be bit too early for this, but definitely we should think about it. Cheers

[mariuszwojcik]

@CESARDELATORRE, earlier you mentioned that below projects will be going through major code changes and we should not convert them to F# yet. Can you update us on the status? I have just migrated MulticlassClassification-GitHubLabeler sample and am looking for next task :)

• eShopDashboardML
• BinaryClassification_CreditCardFraudDetection
• DeepLearning_ImageClassification_TensorFlow
• DeepLearning_TensorFlowEstimator

[CESARDELATORRE]

I would wait some more time until the API is even more stable. Updating all the samples to F# every month means a significant amount of work.
I'm closing this issue. If needed let's create new issues for particular F# topics.

[CESARDELATORRE]

Closing this issue as it is stale