Closed ohltyler closed 3 months ago
Might have missed it in the code but also if its a pretrained model it has the info like dimensions in the model metadata for further steps that need that
Might have missed it in the code but also if its a pretrained model it has the info like dimensions in the model metadata for further steps that need that
Right. I have this in line 196 in common/interfaces
. Now that we have it persisted, we can utilize it dynamically during template construction later on. I still have it hardcoded atm, but there's a TODO tracker for it
Description
This PR adds the ability to select existing models, or select from available pretrained models, in the text embedding processor component. Specifically:
ModelField
field type to process all of the logic around model selection, including only showing valid deployed models, and available pretrained models. Note: the final props and how customizable this field will need to be is TBD. For example, we can customize it to be tailored towards text embedding models vs. other types of models in the futuretoTemplateNodes
to dynamically handle the template node creation based on a pretrained vs. existing/deployed model. If a pretrained model is selected, we add a previous step to register and deploy that model, and use the produced model ID in the downstreamcreate_ingest_pipeline step
. Else, use the existing logic and just use the user/form-supplied model ID directly in a singlecreate_ingest_pipeline
step.Misc:
Flow Framework
breadcrumbDemo video:
screen-capture (29).webm
Issues Resolved
Makes progress on #73
Check List
--signoff
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license. For more information on following Developer Certificate of Origin and signing off your commits, please check here.