I've seen a few tutorials on Loras and I would always run into a problem along the way but your method is very smooth and easy to follow and have come out pretty good. The only thing I've struggled with is training at 768 size instead of 512, it seems to cause errors so I've just settled with 512 for now.
9 місяців тому+2
I am getting stuck in the Sample Lora (part 3): the log ends with 'ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor''. Seems to be an xformers and pytorch issue.. @MalverseAI any idea why? Is the Sample Lora in the link of your description still up to date?
Dang man, I watched other youtube videos & they making me do everything the hard way. This seems like a much smoother method, I'm new to A.i. generation. Most of the time I never get any answers when I ask around about making LORAs.
instead of creating lora for the face, can you create lora for a specific body? because face i can already face-swap. what im looking for is consistent body. And for that, would i get my dataset with photos with the face or just the body?
Hello, I'm getting a "AttributeError: 'ImageDraw' object has no attribute 'multiline_textsize'" error at the last phase of drawing the X/Y chart (11:30) in the video. The plot diagram does not output in Stable Diffusion. A brief google search shows that this may have to do with the Python version?
RuntimeError: Error(s) in loading state_dict for UNet2DConditionModel: size mismatch for conv_in.weight: copying a param with shape torch.Size([320, 9, 3, 3]) from checkpoint, the shape in current model is torch.Size([320, 4, 3, 3]). this is the issue i'm encountering in the second step :(
replace the contents of the text files to the trigger word like "margotrobbie001", "margotrobbienew", etc. make the trigger word unique especially for celebrities because "margot robbie" could be already trained in the model that you are using... You can also change the text file to describe the picture with the person for example "margotrobbie001 wearing pink dress", etc
Sounds like the weight of your lora is not in the correct format. Typically it would read as “” for exg. Keep the “.” In between the numbers “1.4”, or “2.1” this indicates the strength of the lora as the “x/y/z” testing indicates in the tutorial in SD1111.
Typically I just use roop and get some of the best looking faces when combing actor face.. but this probably gives more blended and less photoshop feeling
I dunno what you did at 11:00 in but you already had an image generated in stable diffusion. You dont explain how to add in the prompts your own lora that youve created. This is where I am stuck
Nice video there, I keep getting this error: The model you selected is invalid or corrupted, or couldn't be downloaded. You can use a civitai or huggingface link, or any direct download link." I have used many link to different models but still not working
works fine until the step where I want to open stable diffusion. I copy the prompt etc and then it gives me an error because the GPU on collab has disconnected. Any suggestions?
Hello i'm getting this error curl: (23) Failure writing output to destination cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory python3: can't open file '/content/launch.py': [Errno 2] No such file or directory could you please help
I somehow cannot get past the steps/epochs training step. All I get is "Error. Your total steps are too high. You probably made a mistake. Aborting..." and thats it. I tried to use as many steps what you recommended. I also tried lower number - still getting the same error.
How do i fix this error: Warning: such file or directory 0 225k 0 994 0 0 768 0 0:05:01 0:00:01 0:05:00 768 curl: (23) Failure writing output to destination cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory
Hello, sir. Could you please, if you don't mind, train an actress, let's say Alexandra Daddario original to pose using controlnet or any other tool you might know, to make her sniffing her armpits? Like i wanna see how far one can train SD to show an output like that?
I'm currently on the Lora training (after creating the dataset with 31 images, that's gone well), but the result of running gives me this: KeyError: 'time_embed.0.weight
How to take the final step, but in foocus? Is it possible? Coolab disconnects all the time.
I've seen a few tutorials on Loras and I would always run into a problem along the way but your method is very smooth and easy to follow and have come out pretty good. The only thing I've struggled with is training at 768 size instead of 512, it seems to cause errors so I've just settled with 512 for now.
I am getting stuck in the Sample Lora (part 3): the log ends with 'ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor''. Seems to be an xformers and pytorch issue.. @MalverseAI any idea why? Is the Sample Lora in the link of your description still up to date?
I have the same problem
same problem
!pip install torchvision==0.15.0
@@mdken2198 this solved ur issue, can share ur code?
Dang man, I watched other youtube videos & they making me do everything the hard way. This seems like a much smoother method, I'm new to A.i. generation. Most of the time I never get any answers when I ask around about making LORAs.
Hey , when i writed x value and y value i see RuntimeError: Prompt S/R did not find -0.5 in prompt or negative prompt. how i can correct ? Thank you
Did you find decision?
i had the same error
instead of creating lora for the face, can you create lora for a specific body? because face i can already face-swap. what im looking for is consistent body. And for that, would i get my dataset with photos with the face or just the body?
No one knows?!
I'd say the face can be on the dataset but you should fine tune your caption
Probably your prompt
maybe use a body weight slider Lora?
I'm not getting the link of Gradia, what's the fix?
I Big understand, where u got sampleLora??!!!!!
What is the best configuration for "Anime Characters"? SD 1.5? Tags? Steps? Épochs? (I have 12 images)
Hello, I'm getting a "AttributeError: 'ImageDraw' object has no attribute 'multiline_textsize'" error at the last phase of drawing the X/Y chart (11:30) in the video. The plot diagram does not output in Stable Diffusion. A brief google search shows that this may have to do with the Python version?
hmmm, resolved this by unchecking the "Draw legend" box. I don't really know what this option does though...
great tutorial but please make soundtrack softer preferably instrumental since it clashes with your voice and it is so important to hear your details
Could I create a Lore with a character made by AI too ? If it could be possible can I take this Lora and use it on different Models too ?
Yes
RuntimeError: Error(s) in loading state_dict for UNet2DConditionModel:
size mismatch for conv_in.weight: copying a param with shape torch.Size([320, 9, 3, 3]) from checkpoint, the shape in current model is torch.Size([320, 4, 3, 3]).
this is the issue i'm encountering in the second step :(
the lora i want to make is trained using pony model . Will your sample testing notebook work with pony too ?
Hello how do you find the trigger word for each of those Lora's
Did you find the answer? I’d like to know as well
replace the contents of the text files to the trigger word like "margotrobbie001", "margotrobbienew", etc. make the trigger word unique especially for celebrities because "margot robbie" could be already trained in the model that you are using... You can also change the text file to describe the picture with the person for example "margotrobbie001 wearing pink dress", etc
@@thejamesarnold ty fam 👍🏼
Good and simple. I wish you'd explain triggerwords and caption more, though
HI people,
Did anyone have this error?
RuntimeError: Prompt S/R did not find -05 in prompt or negative prompt.
Sounds like the weight of your lora is not in the correct format. Typically it would read as “” for exg. Keep the “.” In between the numbers “1.4”, or “2.1” this indicates the strength of the lora as the “x/y/z” testing indicates in the tutorial in SD1111.
Typically I just use roop and get some of the best looking faces when combing actor face.. but this probably gives more blended and less photoshop feeling
My colab also gets disconnected
getting ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'
any help?
Is cropping to 512x512 mandatory?
i was gonna ask the same question
Thank you so much for this video. You're a life saver.
I am getting error at middle you are using colab pro?
@@raistar72 free
@@СлесарьЛеха what is the reason for error "balance: 0"
What part of stealing someone's identity is so vital to your existence that you consider this tutorial a life saver? Lmao...
I dunno what you did at 11:00 in but you already had an image generated in stable diffusion. You dont explain how to add in the prompts your own lora that youve created. This is where I am stuck
This step is just for testing the epochs and weights. Just write your own prompts
Nice video there, I keep getting this error: The model you selected is invalid or corrupted, or couldn't be downloaded. You can use a civitai or huggingface link, or any direct download link."
I have used many link to different models but still not working
Same issue
Amazing video
Google colab giving me gpu backnd error after downloding model is it bug?
TYSM for sharing it was very helpful.
No problem 😊
I am getting error at middle you are using colab pro?
How do you load the notebook it doesnt come up on mine
same. Anyone?
can i try all 10 at once?
i cant install kohya bc my python version is 3.12 :(
works fine until the step where I want to open stable diffusion. I copy the prompt etc and then it gives me an error because the GPU on collab has disconnected. Any suggestions?
Google have started banning the use of sd web ui
After clicking the Gradio link it says No interface is running right now..... How do i fix this??
It worked the second time because I didnt stop the colab like you suggested.
Thank you that works
Pro colab or normal google colab
i use pro colab @@raistar72
the best the best the bestttttt i love you
I am getting error at middle you are using colab pro?
sample doesn't works, Collab stops it
Hello
i'm getting this error
curl: (23) Failure writing output to destination
cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory
cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory
cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory
cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory
cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory
cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory
python3: can't open file '/content/launch.py': [Errno 2] No such file or directory
could you please help
same
I somehow cannot get past the steps/epochs training step. All I get is "Error. Your total steps are too high. You probably made a mistake. Aborting..." and thats it. I tried to use as many steps what you recommended. I also tried lower number - still getting the same error.
..nevermind, i had to change epochs to steps in drop-down menu, my bad!
when I put the numbers for the x type prompt it gives me an error that there isnt (number) in the prompt
check on your drive which are the correct numbers, or just test 1 by 1 the Loras
colab doesn't support gradio webui anymore but its still a good tutorial
What alternatives should I use?
@@popaname check diffusers
what does this mean?
@@popaname diffusers
And what do I do when the Model is corrupted?
When downloading through a url, models can not download correctly, to fix this manually download the model locally and then upload to your notebook.
Can you please help me with making one
Do I need a better google collab plan? I was in the a1111 ui and got disconnected from collab saying something about credit and time
Free plan doesnt work
@@Emsyaz ok ty,
I run into an issue. how do i install gradio?
google is banning gradio and other web UI.
runtime disconnected
when i run simple lora ipynb it does not complete and end the runtime any solution for that ?
Following
Now this helped me out. Big thanks!
No problem!
How do i fix this error:
Warning: such file or directory
0 225k 0 994 0 0 768 0 0:05:01 0:00:01 0:05:00 768
curl: (23) Failure writing output to destination
cp: cannot create regular file '/content/microsoftexcel/models/Lora': No such file or directory
I am getting error at middle you are using colab pro?
Why my google colab keep disconeting”
You need pro
@@raghulmanikandan If i upgrade to pro, then I can run the sample_lora? thank you
GPU DATA ID NOT FREE!!!
vous n'avez pas lu mon message , donc je le supprime !
"Easy" lol...
Well, it is ... Lol
free? last time i test something. space was not enought xD. we can do FREE on local. ldets not give our files to google?
all this effort, it didn't catch the essense of the person at all though, thanks though
Hello, sir. Could you please, if you don't mind, train an actress, let's say Alexandra Daddario original to pose using controlnet or any other tool you might know, to make her sniffing her armpits? Like i wanna see how far one can train SD to show an output like that?
everyone has a problems and you dont any respond. DELETE Video. WHAT a waste of a time!
You are going through the tutorial too fast. I can't keep up with you!
I'm currently on the Lora training (after creating the dataset with 31 images, that's gone well), but the result of running gives me this: KeyError: 'time_embed.0.weight
Anyone getting :
"ModuleNotFoundError:
No module named 'torchvision.transforms.functional_tensor'" ?
me too, i tried with different versions of torchvision , didnt work
colab ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'
same error , u found a way around ?
no such file direcotory appears to me i will try later
it worked