Vertex AI- train model from checkpoint?

[bharathsrikanth]

Hi,

I'm looking to use Vertex AI for an entity extraction use case where we have a small amount of training data available now, but expect to have a more data available from production at a future time. As a result, I'm looking to train a version of the model now, with the goal of improving it with more data later.

Is it possible to train the new version of the model using the checkpoint obtained from the first model as a starting point? More generally, are we able to kick off training using a pre-trained checkpoint, or are we required to start training from scratch every time we'd like to create an improved model?

[Qasim_Taleb]

Hello
It is possible to train the new version of the module using the Checkpoint obtained from the first module as a starting point

TEST
احصل على Outlook for Android<>

[bharathsrikanth]

Thank you! Is it possible to point me to some documentation on how to go about doing this?