high quest
Release 2.2 is on its way! Before we release, the team is looking for volunteers to try out the source code installer. See announcement and install instructions here: https://github.com/invoke-ai/InvokeAI/discussions/1628
The binary installer tests were not promising, but I have more hope for the source installer.
split furnace
right off the bat
File "/Users/user/software/AI/Invoke_source/install.sh", line 19
Linux*) OS_NAME="linux";;
^
SyntaxError: unmatched ')'
Oh gee. How did you try to launch the script? From the command line try /bin/bash install.sh
narrow dew
So far window installer seems to be installing so 🤞
Worked yay! For windows at least 🙂
high quest
Try out the new web interface if you haven't already been playing with it. It's terrific.
narrow dew
Yeah have it up now anything you want me to take a look at in particular?
high quest
Try the Unified Canvas (icon #3) from the top. See if you can figure out how to inpaint and outpaint. There are unformatted docs here: https://github.com/invoke-ai/InvokeAI/blob/development/docs/features/UNIFIED_CANVAS.md but I'm interested if you can find the features without hints.
narrow dew
kk I won't put anything up unless I can sort it out
high quest
You might also want to play with the concepts library. https://github.com/invoke-ai/InvokeAI/blob/development/docs/features/CONCEPTS.md It lets you load a variety of styles, subjects and artists from the Hugging Face concepts library. You simply type the concept in <angle-brackets> and the concept will be applied to your generation.
split furnace
Oh gee. How did you try to launch the script? From the command line try `/bin/b...
ok, sorry that was dumb
Installed and launched
The problem is
** "stable-diffusion-1.5" is not a known model name. Please check your models.yaml file
It's clearly ignoring .invokeai despite
Initialization file /Users/user/.invokeai found. Loading...
high quest
Did configure_invokeai.py run (loading models, etc)? Does --root in the .invokeai file point to the right place?
split furnace
yeah it did the configure, and did download a whole lot of stuff I already have. I've skipped downloading SD models because they are in another place
the root points where it should because the manual development branch install works fine
high quest
What platform are you on? Mac or Linux? And could you please paste in the invoke.py startup messages?
split furnace
Mac. Btw do I manually start the conda end in this mode?
Mac
here's the whole run
~/s/A/Invoke_source (main)> ./invoke.sh
Do you want to generate images using the
1. command-line
2. browser-based UI
3. open the developer console
Please enter 1, 2, or 3: 2
Starting the InvokeAI browser-based UI..
* Initializing, be patient...
>> Initialization file /Users/user/.invokeai found. Loading...
Compiling and loading c extensions from "/Users/user/miniconda3/envs/invokeai/lib/python3.10/site-packages/patchmatch".
make: *** No rule to make target `clean'. Stop.
>> Patchmatch not loaded, please see https://github.com/invoke-ai/InvokeAI/blob/patchmatch-install-docs/docs/installation/INSTALL_PATCHMATCH.md
>> GFPGAN Initialized
## NOT FOUND: CodeFormer model not found at ./models/codeformer/codeformer.pth
>> CodeFormer Disabled
>> ESRGAN Initialized
>> Using device_type mps
** "stable-diffusion-1.5" is not a known model name. Please check your models.yaml file
* --web was specified, starting web server...
>> Initialization file /Users/user/.invokeai found. Loading...
>> Started Invoke AI Web Server!
>> Default host address now 127.0.0.1 (localhost). Use --host 0.0.0.0 to bind any address.
>> Point your browser at http://127.0.0.1:9090
>> System config requested
>> Image generation requested: {'prompt': 'galadriel test', 'iterations': 1, 'steps': 20, 'cfg_scale': 13, 'threshold': 0, 'perlin': 0, 'height': 512, 'width': 512, 'sampler_name': 'ddim', 'seed': 4156000351, 'progress_images': False, 'progress_latents': True, 'save_intermediates': 5, 'generation_mode': 'txt2img', 'init_mask': '...', 'seamless': False, 'hires_fix': False, 'variation_amount': 0}
ESRGAN parameters: False
Facetool parameters: False
{'prompt': 'galadriel test', 'iterations': 1, 'steps': 20, 'cfg_scale': 13, 'threshold': 0, 'perlin': 0, 'height': 512, 'width': 512, 'sampler_name': 'ddim', 'seed': 4156000351, 'progress_images': False, 'progress_latents': True, 'save_intermediates': 5, 'generation_mode': 'txt2img', 'init_mask': '', 'seamless': False, 'hires_fix': False, 'variation_amount': 0}
** "stable-diffusion-1.5" is not a known model name. Please check your models.yaml file
'NoneType' object has no attribute 'modules'
Traceback (most recent call last):
File "/Users/user/software/AI/InvokeAI/backend/invoke_ai_web_server.py", line 1116, in generate_images
self.generate.prompt2image(
File "/Users/user/software/AI/InvokeAI/ldm/generate.py", line 405, in prompt2image
configure_model_padding(model, seamless, seamless_axes)
File "/Users/user/software/AI/InvokeAI/ldm/invoke/seamless.py", line 15, in configure_model_padding
for m in model.modules():
AttributeError: 'NoneType' object has no attribute 'modules'
The error messages are consistent with not being able to find models.yaml just like you say. Could you have a look at the file and see if it is still there and intact?
You don't need to start conda manually in this mode. The wrapper shell script starts it for you.
split furnace
it is definitely in invoke0 folder and like I say the old-school install addresses it properly
I don't know why it has the idea to look for stable-diffusion-1.5 – I don't have that line in my models.yaml (named it Sd15 for brevity), so it must be looking elsewhere? In the examples file?
high quest
Oh, ok. This explains it. Do you have any of the stanzas marked default: true? If it doesn't see a default, it tries a hard-coded value. I will fix it so that it just takes the first stanza.
You can also launch with --model=Sd15
split furnace
No dice. In models.yaml I do specify default: true for Sd15.
Btw the entirety of my .invokeai is as follows
# InvokeAI initialization file
# Put frequently-used startup commands here, one or more per line
# Examples:
# --web --host=0.0.0.0
# --steps 20
# -Ak_euler_a -C10.0
--root_dir /Users/user/software/AI/Invoke0
--outdir /Users/user/software/AI/IMG_out
-s 20
so I'm positive this install does not actually follow from .invokeai to the configs
high quest
Ok, I'll investigate. Thanks for your testing!
soft thicket
Installer worked fine on Windows 10, but when I try to outpaint a small selection and have the Images set to more than 1 it succeeds on the first image generation, but fails on the next with the error: ``` Traceback (most recent call last):
File "e:\test\canvas\ldm\generate.py", line 486, in prompt2image
results = generator.generate(
File "e:\test\canvas\ldm\invoke\generator\base.py", line 91, in generate
image = make_image(x_T)
File "E:\test\Canvas\installer_files\env\envs\invokeai\lib\site-packages\torch\autograd\grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
File "e:\test\canvas\ldm\invoke\generator\inpaint.py", line 295, in make_image
samples = sampler.decode(
File "E:\test\Canvas\installer_files\env\envs\invokeai\lib\site-packages\torch\autograd\grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
File "e:\test\canvas\ldm\models\diffusion\sampler.py", line 363, in decode
x_dec = xdec_orig * mask + (1.0 - mask) * x_dec
RuntimeError: The size of tensor a (16) must match the size of tensor b (64) at non-singleton dimension 3
Could not generate image.```
Setting Scale Before Processing to None stops the error from occurring and it generates all the images
high quest
@uneven gazelle - here is the tensor match size issue I ran into, now seen in the wild
uneven gazelle
Is my fix in there yet?
I don't remember if my PR was against development or merged into @desert vault pr
uneven gazelle
desert vault
Oh
uneven gazelle
Was merged in a few days ago
This sounds different though. I guess maybe iterations don't reset the noise either?
desert vault
This sounds different though. I guess maybe iterations don't reset the noise eit...
grunts
Yep - can recreate with images > 1, while outpainting, when inpaint_width / inpaint_height are not the same as the bounding box dimensions
green sphinx
Can I just copy over the configs/models.yaml ?
If I set the path to the 2.1 dir, it's not going to write in there ?!
copying wont work, paths are not absolute
I try to edit the model.yaml by hand then
Can I test this without downloading or duplicating models ?
narrow dew
I mean you could use the base models that it comes with but if you want all the bells and whistles of any custom models you'd have to copy over your models.
green sphinx
Nevermind, the script asks you for paths...
narrow dew
oh you're still installing gotcha well good enjoy it's great stuff. 🙂
green sphinx
But how am I going to pass the --host flag to the web selection ?
Can't reach the webui without that
narrow dew
once you're done with the imports you just have to relaunch the invoke.bat and choice the UI option 2.
green sphinx
yep, but I need to pass the host address, the default won't do it
narrow dew
Oh that I don't know about just the import stuff cause I have lots of checkpoints but I'm sure someone will help.
high quest
Hi folks: new version of the source installers are now available. Will hopefully correct the bugs people have experienced on the first round: https://github.com/invoke-ai/InvokeAI/discussions/1628
narrow dew
❤️
narrow dew
So ran though the installer and everything installed fine but when I launch with the Webui it just sits on a white screen nothing happens. no error messages though on the command line
soft thicket
So ran though the installer and everything installed fine but when I launch with...
Thought I did something wrong, but same problem. Windows 10
high quest
This sounds unsettlingly familiar. Is there an error message on the console/terminal?
high quest
Windows?
narrow dew
Yes sir
high quest
Look for an update in about 30 minutes.
narrow dew
Copy that
high quest
@narrow dew and other users: Problem should be solved. Run ./update.sh from within the directory.
atomic dew
Can I just run update to get 2.2 or full reinstall?
high quest
Yes, just run update.sh. No need to do a full reinstall. The update will download a web file that was missing earlier today.
high quest
🥳
atomic dew
Did a full install on Win 11 PRO, no errors, no deprecation warnings. Running fine so far. Good job team!
high quest
@green sphinx Hang on a minute, I'm about to release a new version. What OS are you using?
narrow dew
I'm standing by with Windows if you don't have anyone available for that OS 🙂
green sphinx
<@616675935118163968> Hang on a minute, I'm about to release a new version. What...
Debian 11, sure I don't have to redownload and install ?
high quest
We're all set for Windows (this was needed for the binary installler). I need someone with an Intel Mac. It won't break your system at all.
high quest
Debian 11, sure I don't have to redownload and install ?
You might. I'm just preparing the zip file now.
green sphinx
You might. I'm just preparing the zip file now.
Oh, so Linux: https://github.com/invoke-ai/InvokeAI/releases/download/2.2.0-rc1/invokeAI-src-installer-linux.zip isn't what I need ?
I fetched that other installer from 10 hours ago
high quest
Thanks for your patience. Here is the pre-release announcement. At the bottom you'll find the installer zip files: https://github.com/invoke-ai/InvokeAI/releases/tag/2.2.0-rc2
It's the same deal: unzip the files, enter the invokeAI directory and run install.sh or install.bat
atomic dew
I have a Mac Mini with Intel CPU and 64GB of ram, but there is no GPU. Will that help you?
high quest
That ought to work. This compiled file is for machines without a GPU. Are you comfortable with git?
atomic dew
Not really, other than just downloading.
If you have a step-by-step, I can figure it out
high quest
You can follow these directions: https://github.com/invoke-ai/InvokeAI/discussions/1617
When you get to the last step of "git commit and push", don't bother. Just take the py3.10* file that this recipe creates and post it here. I'll pick it up.
And thanks very much for your help!
atomic dew
will do
atomic dew
got this message
ERROR: Could not find a version that satisfies the requirement torchvision==0.13.1 (from -r installer/requirements.in (line 23)) (from versions: 0.1.6, 0.1.7, 0.1.8, 0.1.9, 0.2.0, 0.2.1, 0.2.2, 0.2.2.post2, 0.2.2.post3)
i had to install python and pip
next gazelle
On windows 10, I got this error (source install): #1044860655422943302 message. I had to disable my AV (Kaspersky) to be able to install Micromamba. The rest of the installation can be completed with the AV running.
high quest
ERROR: Could not find a version that satisfies the requirement torchvision==0.13...
I appreciate your trying this. I will try to modify requirements.in file to allow it to compile on the Mac Mini. Are you up to try again?
atomic dew
Yes, of course, anything for the cause 😉
green sphinx
editing and saving the model.yaml and running invoke.sh will say model files are not configured and call the script to download them Nevermind I tried to populate the yaml into the old dir and wasn't paying attention to the setted inovke environment folder setted
astral shore
Hey, fyi I just tried the download link for the Linux 2.2 version at the top of the post, but it seems as though that link is broken now,
