Closed Nikolai10 closed 4 months ago
See diffusers pull request fix(training): lr scheduler doesn't work properly in distributed scenarios
Current experiments are not affected by this issue (as we always set max_train_steps to 50k) -> should be however resolved in future releases.
fixed, see https://github.com/Nikolai10/PerCo/commit/c63142a65d7bf25e40e9ba68e748ac5de9e068ca
See diffusers pull request fix(training): lr scheduler doesn't work properly in distributed scenarios
Current experiments are not affected by this issue (as we always set max_train_steps to 50k) -> should be however resolved in future releases.