SimpleTuner

SimpleTuner copied to clipboard

next release

1

bghira

Dataset random sampling should be adjusted by the size of the datasets. Right now if you run a 200 sample dataset alongside a 2000 sample dataset, the 200 sample dataset...

AmericanPresidentJimmyCarter

When setting `vae_cache_clear_each_epoch: true` in the `multidatabackened.json` file training randomly stops with the stack trace below. Reproduced on multiple machines, different epochs and images reported as the issue. ``` (id=images-1024)...

riffmaster-2001

Training crashing for images that used to work fine, with error: latent shape mismatch: torch.Size([16, 88, 136]) != torch.Size([16, 88, 144])

23

Using latest version of SimpleTuner producing an error: "latent shape mismatch: torch.Size([16, 88, 136]) != torch.Size([16, 88, 144])" Examples: [error_crash_latent_mismatch_01.txt](https://github.com/user-attachments/files/17007226/error_crash_latent_mismatch_01.txt) [error_crash_latent_mismatch_02.txt](https://github.com/user-attachments/files/17007228/error_crash_latent_mismatch_02.txt) Both images used in the past training without issues....

a-l-e-x-d-s-9

bghira

bghira

bghira