intro-to-ml-with-kubeflow / intro-to-ml-with-kubeflow-examples Goto Github PK
View Code? Open in Web Editor NEW[WIP] Examples for the Intro to ML with Kubeflow book
License: Apache License 2.0
[WIP] Examples for the Intro to ML with Kubeflow book
License: Apache License 2.0
The Apache mailing list has changed its interface, and it is not anymore mod_mbox of Apache HTTP, hence url like http://mail-archives.apache.org/mod_mbox/spark-dev/201911.mbox/ajax/thread?0
will cause the error because of /ajax
part.
By removing /ajax
, the url http://mail-archives.apache.org/mod_mbox/spark-dev/201911.mbox/thread?0
mailing list URL redirect to new interface [email protected], November 2019 but it does not provide MBOX format listing, hence cannot extract the MBOX format elements such as FROM, TO, SUBJECT.
The thread ID pattern is now different too, e.g. https://lists.apache.org/thread/hg85hhvt270of8fdrmb62kfvm7rpl96p
.
You can click "home" then "my server" if your stuck there forever.
Since we do set -ex
we might fail mid script and it would be good for folks to notice that doesn't look like what it should.
We can encourage them to read forward in the guide while waiting for it to finish.
When I try the pipeline of chapter 2 locally with minikube v1.18.1 & Docker version 20.10.5, I have the following error:
serve
simple-sci-kit-kf-pipeline-8jgpg-992827095
This step is in Failed state with this message: Error from server (Forbidden): error when creating "/tmp/manifest.yaml": seldondeployments.machinelearning.seldon.io is forbidden: User "system:serviceaccount:kubeflow:pipeline-runner" cannot create resource "seldondeployments" in API group "machinelearning.seldon.io" in the namespace "kubeflow"
cc @rawkintrevo
When I try to use URL specified in book I get
Sorry, /seldon/mnist-classifier/api/v0.1/predictions is not a valid page
Factor out the Azure and IBM setup parts into separate scripts so folks can call them if they change their mind.
Hello!
$ kfctl apply -V -f https://raw.githubusercontent.com/kubeflow/manifests/v1.0-branch/kfdef/kfctl_k8s_istio.v1.0.1.yaml
How to solve issue with one image
Warning Failed 11m (x4 over 13m) kubelet Failed to pul │
│ l image "gcr.io/kfserving/kfserving-controller:0.2.2": rpc error: code = Unk │
│ nown desc = Error response from daemon: unauthorized: You don't have the nee │
│ ded permissions to perform this operation, and you may have invalid credenti │
│ als. To authenticate your request, follow the steps in: https://cloud.google │
│ .com/container-registry/docs/advanced-authentication │
│ Warning Failed 11m (x4 over 13m) kubelet Error: ErrIma │
│ gePull │
│ Warning Failed 5m52s (x29 over 13m) kubelet Error: ImageP │
│ ullBackOff │
│ Normal BackOff 58s (x50 over 13m) kubelet Back-off pull │
│ ing image "gcr.io/kfserving/kfserving-controller:0.2.2" │
I tried different versions, but didn't solve the issue.
Thanks
When evaluating the model, the validation or test set should be used because classifiers may overfit and yield perfect results on the training set.
I would suggest changing
batch_xs, batch_ys = mnist.train.next_batch(1)
to
batch_xs, batch_ys = mnist.test.next_batch(1)
in the section about querying the model.
Starting from Kubeflow 1.3, all components should be deployable using kustomize only. Any automation tooling for deployment on top of the manifests should be maintained externally by distribution owners.
Hence the way used in the book to use kfctl is now obsolete. Also Kubeflow 1.50 and later is now incompatible with kubeflow 1.22 and later. Need. to be careful with which K8S version to use.
Attempting to execute the example is chapter 2
Python 3.6.9
kubeflow 1.2.0
tensorflow (1.14.0)
requests (2.25.1)
numpy (1.19.5)
Please use tf.one_hot on tensors.
Extracting MNIST_data/t10k-images-idx3-ubyte.gz
Extracting MNIST_data/t10k-labels-idx1-ubyte.gz
...truncated output for readability...
/seldon/ml/mnist-classifier/api/v0.1/predictions
{
"status": {
"code": -1,
"info": "Empty json parameter in data",
"reason": "MICROSERVICE_BAD_DATA",
"status": 1
}
}
Have confirmed request is not empty.
request = {"data": {"names": features, "ndarray": data.tolist()}}
@holdenk Hoping you can help
For folks who want to go further
Hello!
I have an issue on step Training and Deploying a Model
localhost:7777
Pipelines -> Upload -> job.yaml
Run
The rusult is:
This step is in Failed state with this message: error: error when creating "/tmp/manifest.yaml": Post https://10.96.0.1:443/apis/machinelearning.seldon.io/v1alpha2/namespaces/kubeflow/seldondeployments: stream error: stream ID 123; INTERNAL_ERROR
My config is:
$ export \
PROFILE=marley-minikube \
CPUS=8 \
MEMORY=30G \
HDD=80G \
DRIVER=docker \
KUBERNETES_VERSION=v1.16.0
==================
Any suggestions to solve this issue?
Hello!
How to create bucket what is needed for RecommenderPipeline?
I have an errror:
Stream closed EOF for kubeflow/recommender-model-update-j7c5g-2195988956 (ma │
│ in) │
│ main DataPublisher for model rebuild. Minio: url - http://minio-service.kube │
│ flow.svc.cluster.local:9000, key - minio, secret - minio123 │
│ main Error reading file recommender/directory.txt from bucket data - The spe │
│ cified bucket does not exist │
│ main Error deleting file recommender/directory.txt from bucket data- The spe │
│ cified bucket does not exist │
│ main Error writing file recommender/directory.txt from bucket data- The spec │
│ ified bucket does not exist │
│ main Exception in thread "main" error occurred │
│ main ErrorResponse(code=NoSuchBucket, message=The specified bucket does not │
│ exist, bucketName=null, objectName=null, resource=/models, requestId=3L137, │
│ hostId=3L137) │
│ main request={method=GET, url=http://minio-service.kubeflow.svc.cluster.loca │
│ l:9000/models?location=, headers=Host: minio-service.kubeflow.svc.cluster.lo │
│ cal:9000 │
│ main User-Agent: MinIO (amd64; amd64) minio-java/dev
Please assist!
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.