|
|
|
|
|
|
by deaux
200 days ago
|
|
|
Not a verified X user, but happy to exchange here or elsewhere. The latency leap is still the same for us. We're on us-west1 but reports are that it's similar on at least us-central1 if not elsewhere. We simply can't use the finetuned models in prod any more due to this, but whenever we run our automated tests with them, including today, the latency is still there. We haven't seen issues on non-finetuned models. |
|
|