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.
Nif you see those cases have them email us or jump in the discord immediately. we can back date and credit too.
NCould be pop up blocker. that terminate button has a modal pop up to confirm termination. Just make sure the pop up blocker isn't active if you have one.
If that isn't the case....let me know what browser you are using so i can test
If that isn't the case....let me know what browser you are using so i can test
Fthis makes sense
Nmight be an ad blocker too
None of those potentially
Fi think you should change it from popup to other way
N@Dr. Furkan Gözükara @furk
NThank you
Fyou are welcome
M
M@Dr. Furkan Gözükara do you know this error when using masks? on one trainer
Mnvm fixed it, had to clear the cache
Fa new feature added
