Teradata / jupyter-demos

16 stars 19 forks source link

AzureML Integration for Electricity Consumption Forecast #222

Open mrconway opened 1 year ago

DougEbel commented 12 months ago

Need export of onxx

ForbiddenDevil commented 10 months ago

Probably require the latest database version. The latest BYOM version is installed in the clearscape platform but the demo still gives errors.

DallasBowden commented 10 months ago

Onxx still giving error. Waiting for a TD version upgrade. Lake 20.00.00.6 New TD image has been built and waiting for a space issue to be resolved.

DallasBowden commented 10 months ago

The current CI image is a newer Vantage with the older libraries.

DallasBowden commented 10 months ago

CI environment has latest vantage, but older version of BYOM. It has 3 but it needs 5.

DallasBowden commented 10 months ago

Need to make sure we are looking at the latest machine.

ForbiddenDevil commented 9 months ago

onnxruntime library on the nodes should be 1.13.1. The current version is 1.8.1

Image

The support for TreeEnsemble model(which is used in this usecase) in onnxruntime came in the version onnxruntime 1.12.0

shilpa-nalkande commented 9 months ago

try installing onnxruntime latest version on node

DallasBowden commented 9 months ago

Still working to install the latest packages.

DallasBowden commented 9 months ago

Still waiting for onnx runtime version on the node.

DougEbel commented 9 months ago

Pratik indicates: Onnxruntime is upgraded ... it has 1.3.1 plus:

Teradata 17.20.03.18 teradataml 17.20.0.6 val 2.2.0.0 byom 5.0.0.0 modelops 7.0.2

ForbiddenDevil commented 9 months ago

Please review: https://github.com/Teradata/jupyter-demos/pull/519

ForbiddenDevil commented 8 months ago

AzureML requires numpy version <=1.23.5 but default numpy version in the system is 1.24.2 and latest version is 1.26.4

Image

Hence, I have downgraded the numpy version to 1.23.5 in the pip installs. The notebook should work now.

Note: To discuss the effect of this downgrading on other notebooks, if the user runs any other notebook after this notebook.

Commit: https://github.com/Teradata/jupyter-demos/pull/519/commits/4f1be4dd393631484a188b74719def38c80b61e8

DallasBowden commented 8 months ago

If we downgrade numpy, will it affect any other notebooks? Best option right now is to downgrade for the notebooks and then install it back during the cleanup step.

ForbiddenDevil commented 8 months ago

Added force reinstallation of numpy==1.24.2 in the cleanup section with a warning.

Commit: f5ca9d40e3ca51509e9b08e44c9ba21dcfe7ba97