Comments (9)
I've fixed the closing quote problem (and similar problems parsing the prompt). Use ' to get a quote mark.
With respect to overwriting files, I've observed this as well. I will replace the original filenaming algorithm with something that never overwrites a file. Does anyone mind if I replace the 00042.png filenames with a timestamp, such as 22Aug22-14:10:11.png
from invokeai.
I'm going to stick with the simple numbers and just add a few leading zeroes for the time being. The problem with the original code is that it counted the number of files in the current directory and named the next file 1+the count. However, if some of the files had been deleted, the count would be wrong and files would get overwritten because the code didn't check whether it clobbered a file. For revisions, I'll append a dot and an additional 3 digits (nobody's gonna do an img2img 999 times, are they?).
In the future one could specify a printf string on the command line to format filenames according to your preference, e.g.
--filename_format="%Y-%d-%p-%S.png"
The various % formats would be replaced with something like p="text of prompt", S=seed, %Y=year, etc.
from invokeai.
this is for all forks. it's a problem of original code
from invokeai.
Problem appears when trying to regenerate an image with the same seed.
Original image was generated with filename 00008.png
Then, if you run the prompt below (this prompt is slightly different from original) - script overwrites 00008.png file with new image and reports that new image has exactly the same seed. This behavior makes experiments (attempt to create variations of image by tweaking its parameters) more difficult compared to SD discord bot that generated new seed for each newly generated image and the most frustrating part is that original image is overwritten.
Example of prompt that causes the problem:
dream> "photo of an elephant reading a book" -s150 -b1 -W512 -H512 -C8.5 -I./init-images/5.png -f0.5 -S1108544598
from invokeai.
What about adding seed number at the end of each image name? Every generated image should have unique seed.
Like 0000000001_S1108544598.png
Also, I propose to increase number of decimals in filename from 5 decimals to at least 10 because people will generate tons of images.
from invokeai.
This could work. I'd also like to number img2img mods so that the basename of the original file is reused, so something like 0005-revision1.png
from invokeai.
I like idea with revisions! For correct file sorting I propose to increase number of digits to at least 3: revision001.png
Does anyone mind if I replace the 00042.png filenames with a timestamp, such as 22Aug22-14:10:11.png
If you decide to use timestamp, I propose to use UNIX format because filenames could be correctly sorted in File Explorer.
Example of UNIX timestamp: 1661213008
From psychological point of view I would stick with simpler numbers like 00020.png, 00021.png
Then people could easily remember that their favorite image is in files "55" and "73". If you go with UNIX timestamp - their two favorite images will be in 1661213138 and 1661214235 and nobody could remember it :)
from invokeai.
is half precision used as default? i think its the full one from what i see in the dream file
from invokeai.
- Filenames now use the following formats:
000010.95183149.png -- Two files produced by the same command (e.g. -n2),
000010.26742632.png -- distinguished by a different seed.
000011.455191342.01.png -- Two files produced by the same command using
000011.455191342.02.png -- a batch size>1 (e.g. -b2). They have the same seed.
000011.4160627868.grid#1-4.png -- a grid of four images (-g); the whole grid can
be regenerated with the indicated key
Yes, half precision is the default. The arg parser's syntax is a little weird.
from invokeai.
Related Issues (20)
- [enhancement]: Compact view for models list
- [bug]: invokeai-model-install.py download from https://civitai.com dont work in Invoke 4.0rc1 HOT 1
- [enhancement]: Add more image data to queue API
- [bug]: run or ckpt model to diffusers format give this error vr 4.0rc1 HOT 5
- [enhancement]: Call saved workflows via API
- [bug]: [4.0.0rc1] using TAESD as VAE throws error HOT 2
- [enhancement]: support non-mmap BLAKE3 hashing HOT 1
- [bug]: Model load events contain non-serializable data HOT 1
- [bug]: [4.0.0rc1]: Recalling LoRa Settings does not remove additional LoRas
- [bug]: submitting trigger phrase form with no values "submits" and breaks UI
- [bug]: ControlNet regression HOT 3
- [bug]: Depth anything preprocesssor error
- [bug]: unable to kill app during model initial registration HOT 1
- [bug]: `clip-vit-large-patch14` not found when converting a model, (CONVERT_MODEL_ROOT is set before the CLI args are applied)
- [bug]: gradient denoising causes artifacts in combination with LCM HOT 6
- [bug]: The image is shifted in canvas mode HOT 1
- [bug]: Black Images NVIDIA T1000 8GB HOT 3
- [bug]: outpainting and inpainting It gives a very bad result v4.0rc2 HOT 11
- [bug]: The aspect ratio setting and preview don't change when generation parameters are loaded from saved model defaults.
- [bug]: Model manager requires web page refresh to show the newly added models. HOT 3
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from invokeai.