S@Dr. Furkan Gözükara I am running your Gradio app from this post: https://www.patreon.com/posts/95759342
I have downloaded some custom models from civitai in one of the notebook cells like so:
Then, in the gradio app, in the custom model path field, I list that file:
When I try to generate the image, I got this error:
Do you know what could be going on there?
I have downloaded some custom models from civitai in one of the notebook cells like so:
!wget "https://civitai.com/api/download/models/125771" -O /kaggle/temp/toon.safetensorsThen, in the gradio app, in the custom model path field, I list that file:
/kaggle/temp/toon.safetensorsWhen I try to generate the image, I got this error:
ValueError: The provided pretrained_model_name_or_path "/kaggle/temp/toon.safetensors" is neither a valid local path nor a valid repo id. Please check the parameter.Do you know what could be going on there?
Fuse this way better
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözükara on Patreon
Fhas auto downloader as well
Fsadly no idea
Fi posted your issue on their official discord channel
Fthis is massed compute support reply
F
N@Dr. Furkan Gözükara
N
NHello, the RAM of my PC is 2GB and the GPU (see the image above)
Scommendable that you are trying but to get anything done really you'll need a dedicated GPU, 2GB is not enough system memory but even if you had 32GB system memory it still wouldn't matter without the GPU, sadly
N@shotgun messiah
NThank you for your answer, what is the dedicated GPU and what is the minimum RAM accepted?
SIt's the graphical processing unit... historically, gaming hardware manufacturers created them for videogames but overtime realized the technology works extremely well for transformers (type of tech used for AI to create output from input), so mainly they are seen in desktop gaming computers (Nvidia), so even if your laptop had a GPU which it probably doesn't, the laptop GPUs aren't sufficient in most use cases
Syour best option until you can get your hands on a desktop PC with a GPU is something like Massed Compute if you're trying to do LoRA / checkpoint training https://massedcompute.com/
Massed ComputeMassed Compute
Massed Compute is a cloud compute provider that serves cutting-edge GPUs without bulky contracts and unnecessary up-sells.
Sand it has the added benefit of being much much much faster any GPU you could feasibly get your hands on in most cases, although the RTX4090s and those types of things are really good since they have so much memory
KThere is no fast-stable-diffusion in runpod anymore. What should i use? pytorch:2.2.0-py3.10-cuda12.1.1-devel-ubuntu22.04 - i can't get access to HTTP:3000 port after "python main.py --listen 127.0.0.1 --port 3000"
KI realized the error, it turns out 127.0.0.1 is only local access, and in order to have access through a proxy, you need to enter 0.0.0.0 - maybe this will help someone
Fyour pc is a patato
Fyou cant run anything
Fuse kaggle notebooks we have or cloud
Ffor ai 32 gb ram and 12 gb RTX 3060 minimum i suggest
Fwhat do you want to install?
NOk I will try what you suggest and then I will give you news
Fif you can afford 2x32 gb ram + rtx 3090 is one of the best config
M@Dr. Furkan Gözükara possible to run one trainer on runpod?
Fif you can find a good ubuntu desktop template yes
Fmaybe i can find i need to look for
M
Myou can also create your own templates with runpod
Fyes you can
Fbut installing necessary libraries is hard
Fi doubt that ubuntu has them
Mahhh so only solution is massed compute for now
Fcurrently yes unless there is a good template
Fby the way massed compute even added crypto payment
Fwhich is great
Mis their h100 reliable ?
Fin which sense?
Fi tested recently and batch size 7 speed was very decent
Myeah their speed, can it go higher than batch 6 wihtout cross attention on massed compute?
Fit does 7
Fwith both text encoder
Fwith xformers
Fno gradient checkpointing
Mi did some further testing and still no xformers training is best
Ftrue
Fbut it increases vram a lot
Nthat is not ideal. running a test right now.