Brandon Strong

fix(gui): invalid chunk/pad range, fix slider minimum

7

librosa.util.exceptions.ParameterError: Invalid hop_length: 0 on 0 chunk seconds Calculated padded input size per channel: (1). Kernel size: (2). Kernel size can't be greater than actual input size on 0 pad...

Allow custom test/val quantity

1

Test and Validation samples are selected randomly in during pre-config step, but only two each are selected. In a large dataset where the input may have large variation this could...

Along with config, the model-in-training's val/train/test.txt should be saved to log for better unified tracking.

feat(main/cli): add gui command for module root entrypoint

4

gui has relative modules, so being able to enter from __main__ would streamline debugger launching.

After setting your preferred config, pressing infer/stop/start will reset to vc defaults, such as re-enabling auto play and other settings.

Generate batch of images inside same latent space(fixed latent noise / seed?)

2

In Auto1111, in the txt2image section you have the ability to generate images within the same latent space. For example, with the seed and prompt static, changing the controlnet driving...

Slow Image Save speed

17

I have a large batch workflow and the WAS image save node is nearly the biggest bottleneck, it's almost slower than my detailer pass. Takes about 15 minutes to save...

### Expected Behavior When selecting loras, search will be able to search inside any nested folders. ### Actual Behavior When selecting loras, search will not search inside nested folders, only...

Certain diffusion methods such as AnimateDiff suffer greatly with prompts of differing token lengths, making it virtually impossible to use popular textual inversions/embeddings in the negative prompt. [Auto1111 has this...

How fast should I expect this to be?

1

Using the default settings and the cogvlm script, processing is quite slow. It's running on my 4090,taking up about 14.4gb, and taking about 10 seconds per image.