Albert Torosyan

Hi @Sharathmk99, The use-case with tb converter you described above totally makes sense. The idea with syncing data with remove in batches seems to be the best option. Unfortunately implementation...

> Hi @alberttorosyan > > Thank you for the response. It makes sense about the option you suggested. In our team, we run many training per day and MLOps guys...

Thanks @ptaejoon for raising the issue! @Hamik25 could you please share additional context regarding this issue? Lets also explore possible approaches to fix.

@mahnerak have you seen such error previously? as I recall you had a setup for NFS with Aim?

@pkubik wrt this: > Is it in general safer to run a single aim server on the cluster and attach my jobs by TCP? `aim server` is used for remote...

Hey @Soptq! Thanks for submitting this issue. Ability to customize Aim is something we are planning to support in the future. Implementing the plugin system will take some time and...

Current design of Run structured data (including `Experiment` and `Tag` data) is not well-suited for remote tracking implementation and several other features. The central `Factory` interface constructs `Run` `Experiment` `Tag`...

facing similar issue with ``` plotly==5.4.0 matplotlib==3.5.0 ``` it seems the reason is `plolty` uses deprecated API which was dropped in mpl 3.5.0: https://matplotlib.org/3.5.1/api/prev_api_changes/api_changes_3.5.0.html#removals

@cylammarco are you waiting for the fix or you managed to find some sort of workaround here?

@cylammarco thinking the same here... It is possible to apply runtime patch until this gets fixed, but I don't love much the idea of using `unittest` module in production code