stable-diffusion-webui
stable-diffusion-webui copied to clipboard
X/Y plot (Y-axis - checkpoint name) don't load weights if [checkpoint name].yaml exist
Describe the bug When using the X/Y plot and the axis-Y (checkpoint name) script does not load the weight if there is *.yaml. In this case, it is NovelAI.yaml.
To Reproduce Steps to reproduce the behavior:
- Go to 'X/Y plot'
- axis-Y - checkpoint name (SD, WD, NovelAI)
- click generate
- SD, WD row generates without problem
- NovelAI with each image generation load standard config (as I understand it)
LatentDiffusion: Running in eps-prediction mode DiffusionWrapper has 859.52 M params. making attention of type 'vanilla' with 512 in_channels Working with z of shape (1, 4, 32, 32) = 4096 dimensions. making attention of type 'vanilla' with 512 in_channels
- NovelAI row use weights from previous row
and weights in reverse order
Expected behavior
(without NovelAI.yaml)
Desktop (please complete the following information):
- OS: Windows
- Browser Vivaldi
- Commit revision acedbe67d2b8a3af99ca3b9a2f809e7a2db285d1 COMMANDLINE_ARGS=--listen --gradio-img2img-tool editor --xformers --deepdanbooru --disable-safe-unpickle
Same problem. Checkpoint with yaml file not loading in x/y plot. 2nd is reloaded instead of 3rd with yaml.
l_SD_V1.4.ckpt: 0it [00:00, ?it/s] Global Step: 470000 Applying cross attention optimization (Doggettx). Weights loaded. 1 out of 3: 100%|██████████████████████████████████████████████████████████████████████| 28/28 [00:03<00:00, 8.92it/s] Loading weights [e1de58a9] from C:\stable-diffusion-webui\models\Stable-diffusion\wd-v1-3-full.ckpt06<00:06, 8.93it/s] Global Step: 683410 Applying cross attention optimization (Doggettx). Weights loaded. 2 out of 3: 100%|██████████████████████████████████████████████████████████████████████| 28/28 [00:03<00:00, 8.82it/s] LatentDiffusion: Running in eps-prediction mode██████████████████ | 56/84 [00:13<00:03, 8.84it/s] DiffusionWrapper has 859.52 M params. making attention of type 'vanilla' with 512 in_channels Working with z of shape (1, 4, 32, 32) = 4096 dimensions. making attention of type 'vanilla' with 512 in_channels Loading weights [e1de58a9] from C:\stable-diffusion-webui\models\Stable-diffusion\wd-v1-3-full.ckpt Global Step: 683410 Applying cross attention optimization (Doggettx). Model loaded. 3 out of 3: 100%|██████████████████████████████████████████████████████████████████████| 28/28 [00:03<00:00, 8.69it/s] Total progress: 100%|██████████████████████████████████████████████████████████████████| 84/84 [00:32<00:00, 2.61it/s]
Commit revision : c57919ea2a8e4a23a05d21f28928e08bbf34c59e set COMMANDLINE_ARGS=--autolaunch --deepdanbooru --no-half-vae OS : Window/Edge
Can confirm same for me
Closing as stale.