https://kedro.org/ logo
Join Slack
Powered by
hey team, when creating an azureml kedro pipeline ...
# plugins-integrations
g
hey team, when creating an azureml kedro pipeline and trying to run kedro azureml run, I'm facing the following error: AttributeError: 'ComputeInstance' object has no attribute 'min_instances' Have you ever faced that? thanks in advance
j
hi @George Drakoulas! could you share the complete traceback to know where the 
ComputeInstance
object comes from? also the Python, kedro, and kedro-azureml versions you're using
g
thanks! I will send it. In general, I can imagine that I cannot use the same following configurations for two different pipeline deployments, right? i) container registry ii) storage account iii) resource group iv) docker image v) computer instance vi) Azure machine learning project I'm asking since in the past I created and executed the kedro-azreml pipeline, but now after creating a second pipeline with the same configurations, I faced this error message.
kind reminder 🙂 thanks
j
kind reminder to send the traceback 🙂
g
ok thanks, so after running the kedro azureml run -s AZURE_SUBSCRIPTION_ID and filling in the Azure Storage Account Key for storage account kedro-azure-storage, I face the following error: Thank you very much in advance
kedro_error.txt
j
I see this comes from 
<http://azure.ai|azure.ai>.ml.MLClient
. maybe an incompatible version? @George Drakoulas could you share the Python, kedro, and kedro-azureml versions you're using?
g
yes sure thanks
Copy code
black~=22.0
flake8>=3.7.9, <5.0
ipython>=7.31.1, <8.0; python_version < '3.8'
ipython~=8.10; python_version >= '3.8'
isort~=5.0
jupyter~=1.0
jupyterlab_server>=2.11.1, <2.16.0
jupyterlab~=3.0, <3.6.0
kedro~=0.18.14
kedro-datasets[pandas.CSVDataset, pandas.ExcelDataset, pandas.ParquetDataset]~=1.0
kedro-telemetry~=0.2.0
kedro-viz~=6.0
nbstripout~=0.4
pytest-cov~=3.0
pytest-mock>=1.7.1, <2.0
pytest~=7.2
traitlets<5.10.0
flaml
scikit-learn~=1.0
catboost
black
kedro-azureml[mlflow]
j
paging @marrrcin
m
Might be API change on the Azure part. Did you create the compute instance first? If you did, please report it as a bug and include the stacktrace https://github.com/getindata/kedro-azureml/issues
g
Ok I could check that again because I think I used the following configurations two times. My question is should I modify those for two different pipeline deployments, right? i) container registry ii) storage account iii) resource group iv) docker image v) computer instance vi) Azure machine learning project I'm asking since in the past I created and executed the kedro-azreml pipeline, but now after creating a second pipeline with the same configurations, I faced this error message.
m
As long as it is one Kedro project you're fine with using the same for multiple pipelines
g
thanks a lot! same for all of those, right: i) container registry ii) storage account iii) resource group iv) docker image v) computer instance vi) Azure machine learning project
in my case, there are two kedro projects. I think that might be the case with the error; I used the same configurations for two projects and due to overwriting I faced this error. So for two different kedro projects which of the following should be different? i) container registry ii) storage account iii) resource group iv) docker image v) computer instance vi) Azure machine learning project (edited) thank you in advance
kind reminder.. thanks in advance
m
Please provide minimal reproduction repository
g
sure, would be great. I have created a kedro-pipeline deployment package with iris data. Once I will push to GitHub I'll send the link
hey I have now created a simple iris-kedro-azureml repository. You can access that via this link: https://github.com/GeorgeDrakoulas/kedro_azurelml_iris/tree/main The kedro project includes two simple pipelines: data_processing and data_science I would like to deploy in azureml and then if it works I would like to create a 3rd pipeline for the inference to create a pck model that will include both pre-processing and data_science steps. The aim is to register then the model in azure ml and create the endpoint with the registered model.
🥳 1
I deployed this kedro project with kedro-azureml (after some small debugging steps) and I still get the same error after dockerizing and pushing the docker image ...
4 Views
Open in Slack
PreviousNext
Powered by