#Ran twice - no dice?
44 messages ยท Page 1 of 1 (latest)
How recently did you download the installer?
Same day. Basically, I built my server, then downloaded the installer and ran it.
I mean that there was an update pushed very recently that fixes this issue
Wondering if you have it. Can you try downloading again?
Ahh, I have 2.2.3 (from yesterday), and 2.2.4 seems to be the next up, however
got it. Thanks, trying shortly
Benefits of working from home. lol Can do multiple things
I think the previous version of invoke is stopping me from changing permissions somehow. I saw a line last night with a command tto uninstall previous version, but now I can't locate it on the website.
What is it?
Killed and rebuilt the OS. All good. There was way more files in the 2.2.4 download than the 2.2.3. So far, so good.
I was once told I have the patience of Job. lol
There was nothing else running on it, so it was easy.
Downloading weights currently
if you're at weights, you're pretty much golden
patience is a virtue
and will earn you pretty pictures
Thing wants me to go to links and accept terms, but unless I'm a moron, those links are not for terms, and it's causing an issue....
and more fun times.....
the links are only there for you to accept the licenses for those models. but also we might need to review that. thanks for mentioning it
the message is unusual, but maybe expected if this is a self-rolled server. is it ubuntu or debian? try
apt update && apt install -y libglib2.0-0 libgl1-mesa-glx
let me know if that works. might need to update Linux installer
It's Debian, and yes, that installed some things. I went through invoke.sh again, downloaded weights and no errors this time.
Also, it appears that everyone is having same issue regarding accepting anything in those links. Someone mentioned that there isn't anymore.
Now, how to launch? It says use this command:
Where is that directory?
just run ./invoke.sh. it says so in the 2nd part of the message.
Oh for crying out loud. Missed that somehow this am. I'm dumb
Okay, had a few extra things to run/install as I went, which is fine, all worked out and good documentation.
Then this.
?
Not sure if it's an issue or not, but going to the IP:9090 is not bringing anything up.
Your initialization is using your cpu instead of gpu
I guess the VM has to less RAM configured, happened to me when running invoke in a container as well, after blowing the available RAM up it was working again.
@oblique remnant if you've got a GPU in that box, make sure you install the nvidia driver appropriate for your system, and nvidia cuda toolkit. Then everything should work automagically
https://developer.nvidia.com/docs/drive/drive-os/latest/linux/sdk/common/topics/install_sdk_manager/Installing_CUDA_Debians_Linux.html references cuda 11.4; try to see if you can find 11.7, because that's what pytorch expects. but it might work regardless
ah actually nvm, here you go. the latest. https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html