kf5i / k3ai-plugins Goto Github PK

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Additional context
Add any other context or screenshots about the feature request here.

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Additional context
Add any other context or screenshots about the feature request here.

Is your feature request related to a problem? Please describe.
Based on kubeflow/website#2267 feedback it would be great to have a simple approach to kfp to help users to start

Describe the solution you'd like
A double approach may be to have a micro-container with:

• Jupyter (as headless server)
• KFP SDK
  So the user may connect to it remotely with its favorite IDE (as long it support remote connection to the container) and use -
• Jupyter
• Install a virtualenv for the user with kfp

Describe alternatives you've considered
User should install and configure KFP itself and we want simplicity not complexity

Additional context
https://www.kubeflow.org/docs/pipelines/sdk/install-sdk/

Is your feature request related to a problem? Please describe.
We should aim to have a two(2) separate list of plugins:

• Core: plugins graduated by the k3ai contribution community team that respect certain standards
• Community: any bundle/group or single plugin added by the community but not yet graduated

Describe the solution you'd like
A core plugin is something that is written in a way that satisfies at least three basic requirements:

1. Support the current list of Clusters option of K3ai
2. Support a single application or a specific bundle (matched to public known use cases)
3. Is not a simple combination of existing plugins (see point 2)

A community plugin is everything that users describe as a bundle or single application and that represents a form of a proposal for the project itself to be later maintained and supported by the K3ai project.

Additional context
A process, detailed in the contributing guidelines, should take in consideration various options before graduate a community plugin to the core like:

• is the plugin manifest using the best approach (ie helm vs. kustomize vs shell) ?
• has the plugin a wide usage or request?
• does the plugin solve a specific use case? Is it widely accepted as such?

Labelling test issue

Describe the bug
We have a bug!

To Reproduce
Steps to reproduce the behavior:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context
Add any other context about the problem here.

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Additional context
Add any other context or screenshots about the feature request here.

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Additional context
Add any other context or screenshots about the feature request here.

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Additional context
Add any other context or screenshots about the feature request here.

Is your feature request related to a problem? Please describe.
Original request from: https://github.com/kf5i/k3ai/issues/23 by @tinuxnet
We should offer Jupyter Notebooks and JupyterLab as part of the deployment.

Describe the solution you'd like
We have 3 main use cases:

1. Single notebooks/lab as a headless server so people may connect to them from their own IDE's
2. Single notebooks/lab as part of a single deployment so folks may use them directly if they don't want to use and external IDE
3. Single notebooks/lab as part of a specific deployment to leverage some specific features (i.e.: notebook + kfp sdk)

Additional context
https://jupyter-notebook.readthedocs.io/en/stable/public_server.html

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Additional context
Add any other context or screenshots about the feature request here.

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Additional context
Add any other context or screenshots about the feature request here.

This is a test 17

test automation column

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Additional context
Add any other context or screenshots about the feature request here.

Is your feature request related to a problem? Please describe.
The original issue opened: https://github.com/kf5i/k3ai/issues/22 by @tinuxnet

Describe the solution you'd like
Could be useful to add support to this component:
https://v0-5.kubeflow.org/docs/components/pytorchserving/
Considering that pytorch is used massively in research and production also.
https://pytorch.org/blog/model-serving-in-pyorch/

we are about to develop a custom reinforcement model and want to use kubeflow to make our model development and training and testing easy.
we would love to use K3ai to develop our reinforcement. Could you please develop plug-in(s) to support platforms like Keras & similar?

Describe the bug
Due to changes in the logic of the upstream Katib project the plugin actually is not able to execute any trial job.

Expected behavior
Experiments and Trials should appear immediately after.

Additional context
I'm working on V3 of k3ai so this will be fixed within that release.

Describe the bug
Issues with kubeflow-pipelines-traefik on current K3s

To Reproduce
Steps to reproduce the behavior:
on running K3s node ( v1.20.2+k3s1 ) ... where simple jupyter-minimal-traefik works

k3ai apply -g kubeflow-pipelines-traefik
output ... (see 3 Warnings w/ v1beta1 context, that conflicts with 1.20 version)

Plugin YAML group-name: kubeflow-pipelines-traefik
Plugin YAML content: [{github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=1.3.0 kustomize} {github.com/kubeflow/pipelines/manifests/kustomize/env/platform-agnostic-pns?ref=1.3.0 kustomize}], name: kubeflow-pipelines
serviceaccount "application" deleted
serviceaccount "argo" deleted
serviceaccount "kubeflow-pipelines-cache" deleted
serviceaccount "kubeflow-pipelines-container-builder" deleted
serviceaccount "kubeflow-pipelines-metadata-writer" deleted
serviceaccount "kubeflow-pipelines-viewer" deleted
serviceaccount "ml-pipeline-persistenceagent" deleted
serviceaccount "ml-pipeline-scheduledworkflow" deleted
serviceaccount "ml-pipeline-ui" deleted
serviceaccount "ml-pipeline-viewer-crd-service-account" deleted
serviceaccount "ml-pipeline-visualizationserver" deleted
serviceaccount "ml-pipeline" deleted
serviceaccount "pipeline-runner" deleted
role.rbac.authorization.k8s.io "application-manager-role" deleted
role.rbac.authorization.k8s.io "argo-role" deleted
role.rbac.authorization.k8s.io "kubeflow-pipelines-cache-deployer-role" deleted
role.rbac.authorization.k8s.io "kubeflow-pipelines-cache-role" deleted
role.rbac.authorization.k8s.io "kubeflow-pipelines-metadata-writer-role" deleted
role.rbac.authorization.k8s.io "ml-pipeline-persistenceagent-role" deleted
role.rbac.authorization.k8s.io "ml-pipeline-scheduledworkflow-role" deleted
role.rbac.authorization.k8s.io "ml-pipeline-ui" deleted
role.rbac.authorization.k8s.io "ml-pipeline-viewer-controller-role" deleted
role.rbac.authorization.k8s.io "ml-pipeline" deleted
role.rbac.authorization.k8s.io "pipeline-runner" deleted
rolebinding.rbac.authorization.k8s.io "application-manager-rolebinding" deleted
rolebinding.rbac.authorization.k8s.io "argo-binding" deleted
rolebinding.rbac.authorization.k8s.io "kubeflow-pipelines-cache-binding" deleted
rolebinding.rbac.authorization.k8s.io "kubeflow-pipelines-cache-deployer-rolebinding" deleted
rolebinding.rbac.authorization.k8s.io "kubeflow-pipelines-metadata-writer-binding" deleted
rolebinding.rbac.authorization.k8s.io "ml-pipeline-persistenceagent-binding" deleted
rolebinding.rbac.authorization.k8s.io "ml-pipeline-scheduledworkflow-binding" deleted
rolebinding.rbac.authorization.k8s.io "ml-pipeline-ui" deleted
rolebinding.rbac.authorization.k8s.io "ml-pipeline-viewer-crd-binding" deleted
rolebinding.rbac.authorization.k8s.io "ml-pipeline" deleted
rolebinding.rbac.authorization.k8s.io "pipeline-runner-binding" deleted
configmap "metadata-grpc-configmap" deleted
configmap "ml-pipeline-ui-configmap" deleted
configmap "pipeline-install-config-m2k6bmc5m7" deleted
configmap "workflow-controller-configmap" deleted
secret "mlpipeline-minio-artifact" deleted
secret "mysql-secret-fd5gktm75t" deleted
service "cache-server" deleted
service "controller-manager-service" deleted
service "metadata-envoy-service" deleted
service "metadata-grpc-service" deleted
service "minio-service" deleted
service "ml-pipeline-ui" deleted
service "ml-pipeline-visualizationserver" deleted
service "ml-pipeline" deleted
service "mysql" deleted
deployment.apps "cache-deployer-deployment" deleted
deployment.apps "cache-server" deleted
deployment.apps "controller-manager" deleted
deployment.apps "metadata-envoy-deployment" deleted
deployment.apps "metadata-grpc-deployment" deleted
deployment.apps "metadata-writer" deleted
deployment.apps "minio" deleted
deployment.apps "ml-pipeline-persistenceagent" deleted
deployment.apps "ml-pipeline-scheduledworkflow" deleted
deployment.apps "ml-pipeline-ui" deleted
deployment.apps "ml-pipeline-viewer-crd" deleted
deployment.apps "ml-pipeline-visualizationserver" deleted
deployment.apps "ml-pipeline" deleted
deployment.apps "mysql" deleted
deployment.apps "workflow-controller" deleted
application.app.k8s.io "pipeline" deleted
Error from server (NotFound): error when deleting "github.com/kubeflow/pipelines/manifests/kustomize/env/platform-agnostic-pns?ref=1.3.0": persistentvolumeclaims "minio-pvc" not found
Error from server (NotFound): error when deleting "github.com/kubeflow/pipelines/manifests/kustomize/env/platform-agnostic-pns?ref=1.3.0": persistentvolumeclaims "mysql-pv-claim" not found
2021/02/22 21:24:08 Error during delete: exit status 1
warning: deleting cluster-scoped resources, not scoped to the provided namespace
namespace "kubeflow" deleted
Warning: apiextensions.k8s.io/v1beta1 CustomResourceDefinition is deprecated in v1.16+, unavailable in v1.22+; use apiextensions.k8s.io/v1 CustomResourceDefinition
customresourcedefinition.apiextensions.k8s.io "applications.app.k8s.io" deleted
customresourcedefinition.apiextensions.k8s.io "clusterworkflowtemplates.argoproj.io" deleted
customresourcedefinition.apiextensions.k8s.io "cronworkflows.argoproj.io" deleted
customresourcedefinition.apiextensions.k8s.io "scheduledworkflows.kubeflow.org" deleted
customresourcedefinition.apiextensions.k8s.io "viewers.kubeflow.org" deleted
customresourcedefinition.apiextensions.k8s.io "workflows.argoproj.io" deleted
customresourcedefinition.apiextensions.k8s.io "workflowtemplates.argoproj.io" deleted
serviceaccount "kubeflow-pipelines-cache-deployer-sa" deleted
clusterrole.rbac.authorization.k8s.io "kubeflow-pipelines-cache-deployer-clusterrole" deleted
clusterrolebinding.rbac.authorization.k8s.io "kubeflow-pipelines-cache-deployer-clusterrolebinding" deleted
Warning: networking.k8s.io/v1beta1 IngressClass is deprecated in v1.19+, unavailable in v1.22+; use networking.k8s.io/v1 IngressClassList
ingressclass.networking.k8s.io "traefik-lb" deleted
Warning: networking.k8s.io/v1beta1 Ingress is deprecated in v1.19+, unavailable in v1.22+; use networking.k8s.io/v1 Ingress
Error from server (NotFound): error when deleting "https://raw.githubusercontent.com/kf5i/k3ai-plugins/main/common/traefik/kubeflow-pipeline-bind.yaml": ingresses.networking.k8s.io "pipeline-ingress" not found
2021/02/22 21:25:04 Error during delete: exit status 1

and then these containers (and sometimes more) never successfully start (and just perpetually CrashLoopBackOff)
metadata-grpc-deployment
ml-pipeline
ml-pipeline-persistenceagen
metadata-writer

Expected behavior
successful deployment

Screenshots
n/a

Additional context
Also, WRT to "k3ai --help" output, figured I could clone the "k3ai-plugins" repo and tweak the noted manifests that might contribute to warnings, but

• k3ai list --repo http://localhost/k3ai-plugins/contents/core (seems like a poor example syntax, since "contents" is not in place)
• k3ai list --repo http://localhost/k3ai-plugins/core (even with tweaked URI that seems to match repo artifacts)
  Can't load cache:error fetching plugins content: cannot load plugins: invalid character '<' looking for beginning of value will use remote
  cannot load plugins: invalid character '<' looking for beginning of value
• k3ai list --repo https://github.com/kf5i/k3ai-plugins/core (same is true for trying to use this repo directly from CLI)
  Can't load cache:error fetching plugins content: cannot load plugins: invalid character 'N' looking for beginning of value will use remote
  cannot load plugins: invalid character '<' looking for beginning of value

Describe the bug
After the deployment of PyTorch operator we observed the following error in the logs:
kubectl logs pytorch-operator-db5d78f97-b4tmx {"filename":"app/server.go:73","level":"info","msg":"EnvKubeflowNamespace not set, use default namespace","time":"2019-08-15T13:16:38Z"} {"filename":"app/server.go:78","level":"info","msg":"[API Version: v1 Version: v0.1.0-alpha Git SHA: Not provided. Go Version: go1.12 Go OS/Arch: linux/amd64]","time":"2019-08-15T13:16:38Z"} W0815 13:16:38.983656 1 client_config.go:552] Neither --kubeconfig nor --master was specified. Using the inClusterConfig. This might not work. {"filename":"pytorch-operator.v1/main.go:33","level":"info","msg":"Setting up client for monitoring on port: 8443","time":"2019-08-15T13:16:38Z"} {"filename":"app/server.go:200","level":"error","msg":"the server could not find the requested resource (get pytorchjobs.kubeflow.org)","time":"2019-08-15T13:16:38Z"} {"filename":"app/server.go:102","level":"info","msg":"CRD doesn't exist. Exiting","time":"2019-08-15T13:16:38Z"}

To Reproduce
Steps to reproduce the behavior:

1. Deply PyTorch operator as per documentation
2. Check the logs of the pod once is running

Expected behavior
We should observe no errors

We should allow a user to choose if he wants to install the pipelines only or an inference server and if so, if want to install it along with the pipelines or standalone to create an edge inference server.

We should use a tag like -- (i.e.: --triton for NVIDIA triton inference server)

As per https://github.com/kf5i/k3ai/issues/3#issuecomment-706486571 we should also evaluate different approaches to how we manage the flags. By now we probably will have to stick to a more simple way of manage plugins

The list should be IMO:

• Tensorflow Serving - ResNet is merged in #3
• Tensorflow Serving - Mnist is in the works
• NVIDIA Triton CPU and GPU is in the works too
• Seldon
• BentoML
• KFServing
• TFX

Feel free to add others to the list.

Is your feature request related to a problem? Please describe.
Support deploy KFServing while deploying pipeline. Also support only deploy KFServing (with new istio installation or using the existing one).

Describe the solution you'd like
AI Scientist may train their model on other platform, but to free their hands on serving environment deployment, it would be great if we can have it be deployed automatically.

Is your feature request related to a problem? Please describe.
The v2 spec uses plugin definitions of the form v2/argo/argo.yaml. The plugin spec file name should be independent of the plugin name. This would make tooling more convention driven and handling of different naming formats of plugin names would be easier.

I am proposing something like v2/argo/k3ai_plugin.yaml but something as simple as v2/argo/plugin.yaml should suffice.

Adding support of TF Serving with MNIST models, this should be called with --plugin_tfs-mnist

once kf5i/k3ai-core#19 will be merged, we have to change the argo file format

Please add Owners and Reviewers to respective files

We have to see if there's an easy way to support the full Kubeflow as a plugin to offer to the community.

Document how to use #3
Add Example to the TF-Serving-Resnet page

Is your feature request related to a problem? Please describe.
We should support h2o community server allowing customers to run it directly on K3ai

Describe the solution you'd like
As per their documentation, it's possible to deploy it over K8's http://docs.h2o.ai/h2o/latest-stable/h2o-docs/cloud-integration/kubernetes.html

Additional context
An initial effort is already existing, I should be able to open PR in a couple of days

As an alternative to the native Argo pipelines we should allow a user to install Tekton pipelines as explained in the KF documentation here:
https://github.com/kubeflow/kfp-tekton

We should do this using a flag like --plugin_tekton

Create a uninstall when in case of WSL setup to cleanup all the information

Should we update to 1.0.3?
https://github.com/kf5i/k3ai-plugins/blob/main/plugin_kfpipelines/install#L54
WDYT? @alfsuse

This is a labelling test