InvokeAI: Tracker for breaking bugs in new webUI
@lstein, @psychedelicious, @Kyle0654, @netsvetaev (apologies if I missed anyone)
Describe your environment These bugs should occur on any platform
Describe the bug
- Reproducer:
python backend/server.py
Traceback (most recent call last):
File "C:\Users\quaestor\Source\stable-diffusion-stable\backend\server.py", line 22, in <module>
from ldm.dream.pngwriter import PngWriter, PromptFormatter
ImportError: cannot import name 'PromptFormatter' from 'ldm.dream.pngwriter' (c:\users\quaestor\source\stable-diffusion-stable\ldm\dream\pngwriter.py)
- Reproducer: attempt to generate an image
Traceback (most recent call last):
File "C:\Users\quaestor\Source\stable-diffusion-stable\backend\server.py", line 375, in generate_images
model.prompt2image(
File "c:\users\quaestor\source\stable-diffusion-stable\ldm\generate.py", line 344, in prompt2image
results = generator.generate(
File "c:\users\quaestor\source\stable-diffusion-stable\ldm\dream\generator\base.py", line 73, in generate
image_callback(image, seed)
File "C:\Users\quaestor\Source\stable-diffusion-stable\backend\server.py", line 364, in image_done
path = save_image(image, all_parameters, result_path, postprocessing=postprocessing)
File "C:\Users\quaestor\Source\stable-diffusion-stable\backend\server.py", line 310, in save_image
path = pngwriter.save_image_and_prompt_to_png(image, command, filename)
TypeError: PngWriter.save_image_and_prompt_to_png() missing 1 required positional argument: 'name'
Additional context This is a tracker for bugs relating to the recent web UI work. Obviously it is under heavy development, so this is not high-priority work.
About this issue
- Original URL
- State: closed
- Created 2 years ago
- Comments: 35 (14 by maintainers)
@pbaylies #732 adds support for textual inversion generated embeddings, thanks for bringing that to my attention and providing the test file.
I’ve just commented in #266 regarding grids. I don’t think image grids have any business being in the backend. If you want a grid of images, maybe the client can do that for you, maybe you just use an image editor. Suggest we move the burden of support for grids to the client.
I may have to! I was looking forward to doing more code cleanup and working on pipelining the postprocessing steps, but instead I’ll have to spend time with my wife and kids.
I’m in venv/cmake/pip hell right now. It doesn’t look good.