Open Anipik opened 5 years ago
Note that dotnet/machinelearning
doesn't use dotnet/arcade
at all just yet. Will it be possible to use just the RemoteExecutor
package without upgrading the rest of the build system to arcade?
Sure, just consume the dotnet-core feed: https://github.com/dotnet/arcade#getting-started
Closing this issue as no further updates on the issue have been given. @Anipik feel free to reopen the issue if you can provide more information. Thanks.
Curious why the "need-more-info" label was applied even though no question was asked? This issue shouldn't be closed as it still applies.
Can you please share what the Microsoft.DotNet.RemoteExecutor is and why we should be using it? We have a plan to upgrade to to Arcade. But it is not high priority right now.
Can you please share what the Microsoft.DotNet.RemoteExecutor is
Out of process execution of test code. Necessary when environment variables or other process states need to be mutated.
and why we should be using it
You are already using it: https://github.com/dotnet/machinelearning/blob/449719ccc32e82c505be33926ae93bb7c34d25cb/test/Microsoft.ML.TestFramework/RemoteExecutor.cs#L21. This is an outdated copy from CoreFx. Meanwhile we moved RemoteExecutor to Arcade.
The ask is that you get rid of the copy and just use the nuget feed from Arcade. There's no dependency on "arcadifying" the dependency to consume that package. It's just a nuget package which lives on a feed. Let me know if you need more details.
@mstfbl can we please reopen the issue before it's lost in dev/null?
Dotnet team just added a new package in the arcade repo called Microsoft.DotNet.RemoteExecutor package. https://github.com/dotnet/arcade/pull/2176
We use a lot of code from this project to run our tests. This code was added in https://github.com/dotnet/machinelearning/pull/2200
Although slight modifications were made to the code to handle Microsoft.ML based scenarios. But i think we should be able to use this package to run other tests.
cc @ViktorHofer