Flyte enables production-grade orchestration for machine learning workflows and data processing created to accelerate local workflows to production.

Flyte

Hi, in Flyte 0.18.2 and 0.19.0, the pytorch-operator is got deleted from the chart. Is it expected? I cannot find it in the changelog and Slack.

I think the charts are no longer dependent , can be installed independently 

<https://docs.flyte.org/en/latest/deployment/plugin_setup/k8s/index.html#deployment-plugin-setup-k8s|https://docs.flyte.org/en/latest/deployment/plugin_setup/k8s/index.html#deployment-plugin-setup-k8s>

Yup, previously pytorch-operator was here: <https://github.com/flyteorg/flyte/tree/v0.18.1/charts/flyte/templates/pytorch-operator>

When I tried to upgrade to 0.19.0, Helm will remove this flyte’s pytorch-operator first.

Also, in the doc, we can install spark operator separately, but Flyte 0.19.0 chart still support Spark Operator: <https://github.com/flyteorg/flyte/blob/v0.19.0/charts/flyte/Chart.yaml#L15-L19>

I think idea was to reduce dependencies within flight as we add new plug-ins 

<@U02AVSDQEBV> Currently we have 2 helm chart, One for sandbox and other one used for deploying flyte core component. Please read the release notes <https://github.com/flyteorg/flyte/releases/tag/v0.18.2>

After new changes we will suggest you to use `flyte-core`  helm chart for cloud deployment and `flyte` helm chart is only used for sandbox.

hey <@U02AVSDQEBV> just checking in, did you manage to reinstall the operator?

We are migrating to flyte-core from flyte, and go forward without pytorch and spark operator. Currently there’s no user of them so we fine with omitting them.

When the need arise, we may create new chart that depends to flyte-core and other operators — similar to what flyte chart done.

<@U02AVSDQEBV> that was the intention and you can publish it too