i want to make sure my torch and stuff is on latest versions and with bug fixes, still getting oom e
i want to make sure my torch and stuff is on latest versions and with bug fixes, still getting oom errors on comfyui
Fit will install new version
Fi wouldnt do it
Fthe software owner has to test
Fand he upgrades requirements
Zah ok thanks
Zif they update the requirements.txt file but we already have the package installed, will it upgrade or just see that it's already installed and ignore it?
XTried the same, but you end up with a lot of problems, as the CUDA toolkit, Python, Pytorch and the xformers needs to be working on compatible versions. Stick with 3.10.6 or 3.10.11, and let the script install the version it has (even if it is not the latest).
Fas you become more expert
Fyou can try newer libraries
VIs it worth renting a beefier GPU to do 48GB training? Would it be significantly faster than 24GB on a 4090?
S@Dr. Furkan Gözükara Hi are you planning to create a tutorial of how to train my own text or even a group of pdf to talk with my own data? like dreambooth training but with text. There is anybody that knows or share a tutorial on that subject?
୧How much text do you have?
୧chatGPT allows you to upload text files and I think pdfs too, then you can ask it questions about the text
୧Or ask it to play a role related to the text
Syes but is not that private is uploading to a server do you think that your data is nor compromise?
୧I think they have a way to not share data with openAI, since many companies will have private data
Sthat is with the gpt4 pay plan correct?
୧Yeah, but I haven't looked at how it's done so you keep your data private
୧I don't think it's an issue to upload files in the playground, unless it's very sensitive secrets (and you think you might get sued/fired, etc)
Sfor example books that I bought that I can get some data to ask specific question to solve a particular stuff
୧I don't think openAI cares if you upload stuff like books, especially if you bought them (so you own them)
Sbut chatgpt doesn’t give you a .safetensor file so you can keep your training data
୧You can create an assistant in the open AI playground and keep the files uploaded indefinitely
୧So it's kind of permanent
SThank you so much but I need the gpt 4 payment plan right?
୧You don't even need to buy a plan, just fund your account with $5
୧To get access to gpt4
୧I think platform accounts might be separate from regular chatgpt accounts, not sure if that's still the case
୧୧Read the docs on assistants https://platform.openai.com/docs/assistants/overview
୧Although you don't need to run any code, you can do it via the platform UI
SThank you so much for the advise
୧୧I haven't watched all of it, but it seems an okay for a demo
୧btw we're in the wrong channel, this is about stable diffusion
Fyes many times faster with batch size 4
Fso a6000 much better option
Fthis is also one of my aims
Fand a very useful one
Fbut i didnt have chance yet
F
F
F
F
BHello @Dr. Furkan Gözükara, do you offer a training video guide on how to train Stable Diffusion to apply a particular style to any prompt versus creating a specific model/person each time? For example, creating a model or LoRA to generate an image using a specific drawing/illustration style? Thank you
Also, is the part of your paid Patreon?
Also, is the part of your paid Patreon?
FHello. I am giving training
Fother than my videos
F1 hour is 100$ for individuals. 1 to 1 lecture