#A few thoughts. I don't know if they
1 messages · Page 1 of 1 (latest)
Thanks, Allen.
Any docs for the permissioning API? I didn't see it while browsing. But what you say--train with your personal account, then grant access to the production service account--is what I would ordinarily have expected, so I'd prefer to do that if you have such documentation.
https://ai.google.dev/api/rest/v1beta/tunedModels.permissions says "PaLM", but I believe it is the same configuration.
I'll admit that I haven't done it. (Too many other things to try before wading into authentication chaos.)
Thanks. I'll poke at that a bit. 🙂
Right now I just have fine tuning working (I think) with the prod serviceaccount, which is quite weird of course but, hey, yolo.
Hmm, so, I got a bit sidetracked but back to this:
I was able to automate fine-tuning, but when trying to generate content using the tuned model, I get an HTTP 500.
Code is something like this:
m = genai.GenerativeModel(model_name=mymodel)
print(m.generateContent("12345"))
If I call list_tuned_models(), I see the model is shown with state=<State.ACTIVE: 2>. But if the generateContent call still fails with errors like: google.api_core.exceptions.InternalServerError: 500 An internal error has occurred. Please retry or report in https://developers.generativeai.google/guide/troubleshooting