dxqb

minimum network speed parameters for create_pod

1

community pods can have slow network speeds this PR adds support for the 2 graphQL parameters "minDownload" and "minUpload" to runpod.create_pod

added Skip Save interval before automatic saving starts

6

Automatic saving is used to have multiple checkpoints after training, so you can choose the best one. However, you know in advance that the first x000 steps will not be...

### What happened? On a clean windows install, starting a freshly installed OneTrainer causes the error mentioned above. Can be solved by installing "Visual C++ Redistributable für Visual Studio 2015"....

Question: Flow-DPM-Solver

6

Is there already a generic implementation of your Flow-DPM-Solver, such as for Comfy or Swarm? Would you expect it to work equally well for other flow matching models such as...

new Prodigy optimizer version: signigicantly less VRAM

3

Hey, just a heads-up, there is a new version of the Prodigy optimizer that requires significantly less VRAM. * new hyperparameter slice_p==11: saves 45% of VRAM overhead. There should be...

see above the error message is something about "not enough disk space available"

see above

Multi-GPU support

1

**Multi-GPU support for OneTrainer** This is a draft, but it is intended to be feature-complete and work with all models, optimizations and other parameters OT has to offer. Some basic...

[Feat]: Same timestep used for all validation steps

12

### What happened? In this PR, the validation timestep was fixed to 499: https://github.com/Nerogar/OneTrainer/pull/660 This was to make sure each validation uses the same timesteps, and therefore make the validations...

Implementing the conclusions of this thread: https://github.com/Nerogar/OneTrainer/issues/772 Summarized: - Validation on timestep 500 is not ideal, but hardcoded currently - Choosing validation timesteps from a distribution is not good either,...