Editing Training Templates shouldn't update the base model

[dadmobile]

Or at least that's my personal opinion. :) I realize that changing this would require having a way to select a base model.

The reason this is annoying is that my standard flow for training is:

1. Create a template with default parameters
2. Run a test train
3. Run the new trained model to test it
4. Based on how step 3 goes, update the training template and go back to step 2

But right now you have to do an extra step between steps 3 and 4 to eject your newly trained model and open up your base model or else you can't edit the template (without accidentally doing training inception).

[deep1401]

I think it should right? Based on the idea that you have a central model in each experiment within Transformer Lab and editing the template changes it. But maybe we have should one extra tab in the config to select models?

[dadmobile]

When I was actually trying to iterate on a training task it drove me nuts to have to constantly stop my model and flip back to the base model just so I could update a hyperparameter without resetting the base model.

But maybe the answer should just be workflows?