Instructions to build the repo #62
Comments
|
We'll assume that things are built around https://www.electron.build/configuration/dmg |
|
First run build.sh in stable-diffusion and follow the readme in the electron app. |
|
I ran build.sh which seems like it did a lot of things. However, I am running into an issue when trying to run locally: Any help would be appreciated! |
|
I've noticed that |
|
I'm seeing: ...which looks an awful lot like I'm missing some sort of install step... |
|
@schvenk |
|
Getting there: what's left is the missing |
|
The could you include the |
|
Is anyone able to build this at all? Just curious since no one seems to be asking anymore. Just wondering how those who are sending PRs are even running it locally. |
|
The torch backend is not used anymore. |
|
@divamgupta thank you for the pointer, i did manage to get it to build through tf backend. But for some reason it's just giving me "diffusionbee test image" regardless of what prompt I enter. What might I be doing wrong?
[EDIT] I see the error messages in the console. Looks like it's treating it like an empty prompt |
|
Are you using the |
|
@jehan60188 I tried building using Have you successfully built the electron app on your end? |
|
ModuleNotFoundError: No module named 'tqdm' I get this when I want to run 'diffusionbee_backend' that I compiled. Did ANYONE get this to work?
I installed conda, metal drivers, and everything and still can't compile. Would you share quick steps to help us build it? |
@jehan60188 Instead of using |
I've run into the same error after switching from torch to tf backend. |
Can I know why is torch backend on longer supported? Thanks. |
|
Because the other one is running on Tensorflow and is gpu accelerated I’d
guess.
So the problem isn’t npm build, the fake backend is super useful. You need
to set the PY_SCRIPT variable and point it to the real backend (diffusebee).
This will only work though if you have all the metal dependencies setup
correctly
…On Thu, 29 Sep 2022 at 02:27, odieXin ***@***.***> wrote:
The torch backend is not used anymore.
Try running build.sh from
https://github.com/divamgupta/diffusionbee-stable-diffusion-ui/tree/master/backends/stable_diffusion_tf
And let me know if you get any errors.
Can I know why is torch backend on longer supported? Thanks.
—
Reply to this email directly, view it on GitHub
<#62 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAA3UJHT477I6I5BDENFCELWATPA5ANCNFSM6AAAAAAQPCYE4Y>
.
You are receiving this because you commented.Message ID:
***@***.***
com>
|
Same error for me. It looks like a copy recursion loop issue. |
|
@divamgupta it'd be great if you could provide some help here. it would unlock the ability for others to contribute to the project! |
|
You don’t need to build unless you are ready to ship it. I suggest you use
electron:serve instead and start investigating on how to get the backend
to work.
…On Fri, 30 Sep 2022 at 12:39, Grant Ammons ***@***.***> wrote:
@divamgupta <https://github.com/divamgupta> it'd be great if you could
provide some help here. it would unlock the ability for others to
contribute to the project!
—
Reply to this email directly, view it on GitHub
<#62 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAA3UJGHI75FSZGQL45ZFXTWA27PPANCNFSM6AAAAAAQPCYE4Y>
.
You are receiving this because you commented.Message ID:
***@***.***
com>
|
Have these instructions been posted yet? I see new features have been added to the app and I'm eager to try them, but I can't figure out how to run the app and get any results besides the "sample.png" of the astronaut riding a horse image. |
|
Here are the steps to run DiffusionBee with actual TF backend. Hope this helps.
|
This instruction works great, thanks! |
@divamgupta the build.sh provided creates a binary that only uses CPU. MPS/M1 etc is not enabled. So looking at ~8s/it on an M1 Pro instead of ~1s/it. |
"Error: ENAMETOOLONG: name too long" issue is caused by process.env.BACKEND_BUILD_PATH which should be replaced with right value. ` I have successfully built frontend and backend for win 10. But another issue related to GPU need to be further investigated: stderr: 2022-12-19 20:07:30.717842: I tensorflow/stream_executor/cuda/cuda_dnn.cc:384] Loaded cuDNN version 8600 |
|
Issue fixed. stderr: 2022-12-19 20:07:30.717842: I tensorflow/stream_executor/cuda/cuda_dnn.cc:384] Loaded cuDNN version 8600
|
|
@divamgupta I'd just like to reiterate that there's still a need for clear instructions to build this project. In its current state it's a real hurdle for potential contributors. I love what you did with the app, it's my preferred way to generate images but I've used it enough to encounter some pain points (and bugs) which I'd like to help alleviate and fix. The procedure outlined by @zlich is a starting point, but I have encountered quite a bit more hurdles such as a wrong version of Node.js (I was using 18.X but this project seemingly requires 16.X; it would be nice to add a When the time comes to run the When the time comes to run the Electron app, I get a runtime error: Etc. etc. In essence, it's really tough for someone who isn't well-versed in Python (and its pretty complex package / virtual environment ecosystem) to figure out how to run the app. I would personally contribute to the GUI more than the backend(s), so I'd like the Python stuff to just work, but I'm a bit confused by what's happening at the moment. I don't want to just @divamgupta Surely the exact setup requirements are pretty clear to you (say, if you had to setup your project on a brand-new Mac), but it's quite inscrutable to a lot of potential contributors, especially those who aren't working with Python regularly. It'd be incredible if you could write up a detailed procedure to get the project up and running. One way to make sure you don't miss anything would be to setup the project on a new Mac (new user on the same Mac?) or in a different partition to make sure you don't miss any of the dependencies in the instructions. At least that's how I'd proceed. Thanks in advance :) |
Can someone add the instructions to simply build the repo if they don't want to use the downloader?
The text was updated successfully, but these errors were encountered: