I meant to say, I was offered a little job involving Rerender A Video. heh
I meant to say, I was offered a little job involving Rerender A Video. heh
Fmy auto isntaller working
Fyou can use
FYouTubeSECourses
In this video I share automatic Windows installer for ReRender A Video github repository - open source AI project. The auto installer script is here : https://www.patreon.com/posts/1-click-auto-for-89457537
Auto script is here
https://www.patreon.com/posts/1-click-auto-for-89457537
Official repo is here
https://github.com/williamyang19...
Auto script is here
https://www.patreon.com/posts/1-click-auto-for-89457537
Official repo is here
https://github.com/williamyang19...
MOk great, I'm just getting started so that's good to know.
Fgreat
Yamazing. I will email Udemy about this as well. In meanwhile I am buying your subscription and messaging you from there
Fawesome ty so much
Ysigned up and already sent you the message 
.. looking forward to watching videos today
.. looking forward to watching videos todayFok let me check
NI'm considering fine-tuning Stable Diffusion with 80k clothing images, featuring 5k unique products. I plan to assign a unique token for each apparel item. Is this approach suitable for training? I'm using the EveryDream2 trainer.
AMy gpu have 8 gb vram
M
GitHub
[SIGGRAPH Asia 2023] Rerender A Video: Zero-Shot Text-Guided Video-to-Video Translation - GitHub - williamyang1991/Rerender_A_Video: [SIGGRAPH Asia 2023] Rerender A Video: Zero-Shot Text-Guided Vid...
Fhow many unique apparel item you have?
F8 gb is not good to do training
Fyour best option is kaggle SDXL dreambooth for free
Fye solving their repo isntall hard
Fuse my auto installer
Fit works
NI have 5000 for now. I wanted to experiment on this first.
NHave an L40 GPU with 48gb vram
Fhow are you planning to find 5000 uniqu captions haha :d
NPlanning to add new tokens to the clip tokenizer, found an option to do that.
Finteresting where did you find it?
Fit literally tokenize everything
Npipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16).to("cuda")
pipe.tokenizer.add_tokens(list_of_tokens))
pipe.text_encoder.resize_token_embeddings(len(pipe.tokenizer))
pipe.tokenizer.add_tokens(list_of_tokens))
pipe.text_encoder.resize_token_embeddings(len(pipe.tokenizer))
Nlike this
Fwow you gonna train with diffusers good luck
Nsince everydream2 uses diffusers i can simply put these lines in the pipeline
Fit is a very hard job to get it right
Ni am using everydream2
Fi heard it makes more sense
Fi can also give you private consultation for different approach. but it wont be able to bring you certain product
Fit would be more general one
Flike type of clothing etc
Nthat would be helpful
Nbut is it worth trying this approach?
Fi dont think so
Ffirst of all assigning 5000 unique tokens
Fwould take huge time
Fand you dont know if will work or not
Falso 80k images you really should use multi gpu
Fyou can run on runpod really cheap
For you can wait days :d
Nyeah planning to use multi gpus. I have unique ids for each product so planning to put some random alphabets that is not in the clip vocabulary. I have tried this with 500 products and I got decent results. my concern will the model remember all the tokens when I scale it to 5000
Fwith sufficient number of epochs i think should
PStill not seeing anywhere near a 70% increase.
25% increase, yes.
70%, no.
How are you equating this percentage ratio?
25% increase, yes.
70%, no.
How are you equating this percentage ratio?
Fwhat is your gpu
Fi am installing tensoRT right now on my new model
P4090