tensorflow
Help with `tfx.extensions.google_cloud_ai_platform.Pusher`
Thank you for submitting a TFX documentation issue. Per our GitHub policy, we only address code/doc bugs, performance issues, feature requests, and build/installation issues on GitHub and we welcome external contributions!
The TFX docs are open source! To get involved, please read the documentation contributor guide: https://www.tensorflow.org/community/contribute/docs
URL(s) with the issue:
Please provide a link to the documentation entry, for example:
https://www.tensorflow.org/tfx/api_docs/python/tfx/v1/extensions/google_cloud_ai_platform/Pusher
Description of issue (what needs changing):
This tutorial Vertex AI Training with TFX and Vertex Pipelines showcases how the custom_config should be specified. But it turns out that similar configurations do not transfer to tfx.extensions.google_cloud_ai_platform.Pusher and currently there's no documentation on how that should be specified.
Also, this Pusher component is likely going to deploy the model to the legacy AI Platform and not to Vertex AI. Is there a way we could actually deploy this to Vertex AI?
Clear description:
If the team could provide a minimal example of how the custom_config should be specified when using the tfx.extensions.google_cloud_ai_platform.Pusher component with a clear description of each of the arguments, that would be very helpful.
This will essentially allow us to take a model from Trainer (or Evaluator) component and directly push that to Vertex AI that offers niche support for deployment. If this is not figured out properly then users will have write custom components that may not be straightforward.
@sayakpaul,
Please refer to example Vertex AI Training and Serving code's Write a pipeline definition section. This example showcases Pusher with custom_config argument. Thank you!
Closing this due to inactivity. Please take a look into the answers provided above, feel free to reopen and post your comments(if you still have queries on this). Thank you!