ml-commons
ml-commons copied to clipboard
Support Pretrained Model from on-prem or different cloud provider
Is your feature request related to a problem? Today, pre-trained model endpoints and model meta list endpoint are hardcoded and point to openearch artifactory. When OpenSerach is deployed on prem on in different cloud provider, opensearch artifactory endpoint is not reachable and pre-trained model won't be able to used.
What solution would you like?
- Other cloud provider will mirror opensearch artifactory and host it in their artifactory
- Add a setting to customize the model endpoint and model meta list endpoint
What alternatives have you considered?
- Expose internet connection so that other provider can reach to opensearch artifactory. However it is not security since some time cluster requires OpenSearch cluster deployed in private subnet.
Do you have any additional context? Add any other context or screenshots about the feature request here.
PR to address this issue https://github.com/opensearch-project/ml-commons/pull/2119