Microsoft stopped supporting the MLOps of Recommenders

[miguelgfierro]

Description

We need to disconnect the AzureML tests and find another way to have a healthy repo.

TODO:

• Recover the images of READMEs and notebooks: Replace images with new GitHub source #2254
• Recover the assets/params
• Create a new CPU and Spark pipeline
• Create a new GPU pipeline

In which platform does it happen?

How do we replicate the issue?

Expected behavior (i.e. solution)

Willingness to contribute

• Yes, I can contribute for this issue independently.
• Yes, I can contribute for this issue with guidance from Recommenders community.
• No, I cannot contribute at this time.

Other Comments

[miguelgfierro]

from @loomlike:
some thoughts:
Use free CPU VMs for non-gpu pipelines via github action
Use free GPU kaggle actions for gpu-pipelines (but I guess we should only run smoke-test for GPU tests) https://github.com/marketplace/actions/kaggle-action

[SimonYansenZhao]

We can rent a single GPU machine on Azure or other cloud services, and use GitHub Actions to rewrite the AzureML code and run the tests directly on the GPU machine without using AzureML services which may cost extra money.

[miguelgfierro]

@SimonYansenZhao and I discussed options:

• We are going to try first @loomlike proposal: CPU on GitHub and GPU on Kaggle.
• For storing datasets/images we can't use Google Drive because it doesn't have access in China, perhaps we can use Huggingface datasets?
• Since Miguel had to erase the subscription quickly, we need to review if we have lost some of the content we had.

FYI @anargyri