Can you install Invoke on a virtual GPU like runpod?

[valid zephyr]

I tried following the installation steps, but when it gave me a local link at the end I didn't quite know what to do with it. It won't take me anywhere if I open it with my browser.

I'm a designer dipping his toes on ai art, so the patience is appreciated!

[nocturne snow]

As long as PyTorch can see it, yes

You should use --host 0.0.0.0 if it isn't being hosted on the same machine

and then use the machine's local address

usually 192.168.x.x or name.local

[valid zephyr]

What do you mean by "using --host 0.0.0.0"? do I just type that on the terminal?

thanks for the quick reply btw! 🙂

[magic lantern]

Yep! We’re working on a dockerfile

@hybrid urchin - potential testers above! 🙂

[hybrid urchin]

@valid zephyr could you please try this Runpod template: https://www.runpod.io/console/gpu-secure-cloud?template=vm19ukkycf and let us know how it works out for you? There are a few steps to follow in the README, but should be fairly easy! we're here to help if you need it.

[vivid sundial]

Thanks brodsky, came here looking for this answer, and it worked a charm. Thanks for the tip in the readme to let the CPU die down to connect.

[hybrid urchin]

awesome!

[reef sparrow]

I'm gonna try it, but already looks complicated to me.

[reef sparrow]

I guess that at this point I have to connect via HTTP (Port 9090) but I got a problem

[reef sparrow]

Any help please?

[reef sparrow]

Hi, a bit off patience with me... Is there a way to upload from Google Drive to Runpod the models I have? I used AUTOMATIC1111 and I did with Pitorch, but with INVOKE I don´t have the chance. Please help, I have an important presentation and I would like to show why INVOKE is the way to go

[hybrid urchin]

@reef sparrow , let's get a few things in order. Did you make sure to follow the README? At a high level:

1. you first start the pod, it will pause and wait.
2. You use SSH to connect to it and run the configure_invokeai.py script.
3. You edit pod config, commenting out one line of the Docker script for the pod.
4. The pod restarts, and then you can connect over HTTP.
   Can you make sure you follow these steps? this is the only way to do this right now. It will be improved.

[reef sparrow]

Thank you, I reach this point

[hybrid urchin]

regarding Google Drive sync: I don't think Runpod offers that. You can sync with several different cloud storage backends, but Google Drive isn't one of them

[reef sparrow]

And I run properly your amazing tool, the problem now is how to send the custom model to Runpod

I´m investigating with Backblaze, but for an idiot like me it takes time

I said Google Drive because I do the same procedure with AUTOMATIC1111 but with Pitorch

[hybrid urchin]

ah, i see. Glad it's working! so, what you could do, actually.... (with Backblaze or S3):

• sync your /workspace directory up to cloud storage
• upload your custom model .ckpt to models/ldm/stable-diffusion-v1/ inside the workspace cloud copy
• download the models.yaml file from configs/, and edit it locally. Use the examples in the file to add another model, and point it to your custom model .ckpt
• upload the models.yaml back to the storage
• at this point you have a complete /workspace which you can sync back to your pod
• restart the pod. and... .profit? 😅

this should work. at least I don't see a reason why it wouldn't. 🤔

[reef sparrow]

ok, just with 4 lines you gave me homework for 3 days... but thank you

[hybrid urchin]

we're here to help! 😄 you can do it! 😄

[reef sparrow]

👍