Enrico Minack comments

Results 653 comments of


                                            Enrico Minack

Test PTL 1.6.3

Please change this for macOS tests as well: https://github.com/horovod/horovod/blob/master/.github/gen-workflow-ci.py#L458 https://github.com/horovod/horovod/blob/master/.github/gen-workflow-ci.py#L468

horovod build fails with oneccl and tensorflow

This error is very generic, is there something more insightful in the log? Do you have a sequence of commands to reproduce this?

[CI] Upgrade PySpark and Torch

Thanks for approval, good point re macOS. I have upgraded where possible. Waiting for CI to complete.

Fix gradient and var name mismatch when num_groups > 0

@maxhgerlach @Infi-zc what is the status of this PR?

Migrate distutils to packaging.version

Thanks for pointing this out. Would you mind creating a PR for that?

Horovod cannot be installed together with ML frameworks with pip>=22.0

The simplest way is to install the ML frameworks in a separate step first (as we do it in our `Dockerfile.test.?pu`), and then run `pip install horovod[FRAMEWORK]` to get framework-specific...

Horovod0.24.2 +pytorch 1.10.2 throw unexpected fault

I don't quite understand your last message. Did you find a solution or not? What else do you need?

Allow Retry class to determine to retry based on response

@sethmlarson to be clear: this is not intended for 2.0 but the next 1.x release. So I presume this should target branch `1.26.x`, right?

Allow Retry class to determine to retry based on response

So no quick fix. And a non-breaking fix is not required. What is the time horizon of the 2.0 release.

Allow Retry class to determine to retry based on response

Should `is_retry` be marked as deprecated? Or shall both methods co-exist in the future?