New UbiOps features May 2022

UbiOps release news – version 2.16.0

On the 12th of May 2022 we have released new functionality and made improvements to the UbiOps SaaS product. We took your feedback to heart and included some often requested features, as well as general speed and stability improvements. An overview of the changes with accompanying screenshots can be found on this page.

 

Create your free account

✔️ Stability and speed improvements & removal of deployment mode

We made a lot of changes under the hood to increase the speed and improve the stability of UbiOps. One of these changes allowed us to support both batch and express requests for any deployment or pipeline again. Therefore we could take out the deployment modes. Everything will continue to work the same in terms of creating requests, but you will have more freedom and you won’t need to think about what deployment mode you need anymore.

 

✔️ Aggregated requests overviews

We have added an API endpoint to the UbiOps API for retrieving all requests for a given project. This request overview is also visualized in the UbiOps WebApp in the form of a table that can be found under Monitoring > Requests. We added this overview so you can quickly see what is happening in your project.

The aggregated request overview also allows you to immediately check out failed requests in your project, to find out which deployment is causing the problem.

  

We also added a download button with which you can download your requests as a CSV or JSON file. This also works for the request overview tables for individual deployments or pipelines.

 

✔️ Miscellaneous improvements

We added several minor improvements as well:

  • It is now possible to manually trigger a request schedule in the UbiOps WebApp with the “Run now” button. This can be useful for testing purposes.
  • We added the minimum and maximum instances setting to the deployment version overview tables.
  • Warning level logs are now mapped to info log level in UbiOps. Previously they were mapped to the error log level
  • When a pipeline object returns a list of dicts, as opposed to a single dict, each dict will trigger a separate request to the next object in the pipeline. Previously the entire list was tracked as a single request.

 

✔️ Note of warning when using GPUs

For versions that are using ubiops.yaml to install a custom version of CUDA framework, we noticed that NVIDIA has rotated the repository keys.

See https://developer.nvidia.com/blog/updating-the-cuda-linux-gpg-repository-key/. E.g. if you are using the repositories https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64 then you need to add another key to ubiops.yaml

apt:
  keys:
    urls:
      ...
      - https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/3bf863cc.pub

Do you need support?

If you have any questions, do not hesitate to reach out to us via the support portal, or phone. And don’t forget to join us on Slack

Latest news

Turn your AI & ML models into powerful services with UbiOps