How can it be corrupted? I just did your simple tutorial installation for kohya on runpod
How can it be corrupted? I just did your simple tutorial installation for kohya on runpod
Fwhich model you downloaded?
POh I see okay
Fto ensure load model in web ui and generate image
Fthat is what i do
Fto not waste time
POkay will try that
P@Dr. Furkan Gözükara do you know why it won’t provide the link when I run gui.bat?
PI managed to get the link after several tries but then the same thing happens when I click “start training”. Here are the types of errors I receive
P16:02:36-949749 INFO Epoch: 8
16:02:36-949749 INFO Max train steps: 1600
16:02:36-949749 INFO stop_text_encoder_training = 0
16:02:36-949749 INFO lr_warmup_steps = 0
16:02:36-949749 INFO Saving training config to
C:/Users/RENDA/Pictures/loraface2output\model\loraface2_20240522-160236.json...
16:02:36-949749 INFO Executing command: C:\kohya_ss\venv\Scripts\accelerate.EXE launch --dynamo_backend no
--dynamo_mode default --mixed_precision bf16 --num_processes 1 --num_machines 1
--num_cpu_threads_per_process 2 C:/kohya_ss/sd-scripts/sdxl_train_network.py --config_file
C:/Users/RENDA/Pictures/loraface2output\model/config_lora-20240522-160236.toml
16:02:36-966750 INFO Command executed.
16:02:41-172981 INFO Training has ended.
16:02:36-949749 INFO Max train steps: 1600
16:02:36-949749 INFO stop_text_encoder_training = 0
16:02:36-949749 INFO lr_warmup_steps = 0
16:02:36-949749 INFO Saving training config to
C:/Users/RENDA/Pictures/loraface2output\model\loraface2_20240522-160236.json...
16:02:36-949749 INFO Executing command: C:\kohya_ss\venv\Scripts\accelerate.EXE launch --dynamo_backend no
--dynamo_mode default --mixed_precision bf16 --num_processes 1 --num_machines 1
--num_cpu_threads_per_process 2 C:/kohya_ss/sd-scripts/sdxl_train_network.py --config_file
C:/Users/RENDA/Pictures/loraface2output\model/config_lora-20240522-160236.toml
16:02:36-966750 INFO Command executed.
16:02:41-172981 INFO Training has ended.
PFile "C:\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 637, in simple_launcher
raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['C:\kohya_ss\venv\Scripts\python.exe', 'C:/kohya_ss/sd-scripts/sdxl_train_network.py', '--config_file', 'C:/Users/RENDA/Pictures/loraface2output\model/config_lora-20240522-160332.toml']' returned non-zero exit status 3221225477.
16:03:42-841621 INFO Training has ended.
raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['C:\kohya_ss\venv\Scripts\python.exe', 'C:/kohya_ss/sd-scripts/sdxl_train_network.py', '--config_file', 'C:/Users/RENDA/Pictures/loraface2output\model/config_lora-20240522-160332.toml']' returned non-zero exit status 3221225477.
16:03:42-841621 INFO Training has ended.
Fyou dont have git?
Fi bet related to your folders
FC:/Users/RENDA/Pictures/
Fdont use users renda pictures
Fuse like c:/ai_testing
PHmm but the issues occur even before I reach the stage of putting in the path to images. I’ll check the git.
What’s wrong with the current path? I have just stored the data preparation folder in the pictures
What’s wrong with the current path? I have just stored the data preparation folder in the pictures
Sdoes dreambooth lora is capable of style finetuning ?
Fi dont know try it
Fit doesnt show any specific error
Fyes of course
NDo you happen to know what version you are experimenting with? I haven't used tensorrt before but can play around with it a bit too
Fi tried since first day. never worked. i tried forked repo as well
Nim taking a look at this
Fwhichj works on runpod windows
FGitHub
TensorRT Extension for Stable Diffusion Web UI. Contribute to NVIDIA/Stable-Diffusion-WebUI-TensorRT development by creating an account on GitHub.
Nthanks!
FGitHub
TensorRT Extension for Stable Diffusion Web UI. Contribute to andrewtvuong/Stable-Diffusion-WebUI-TensorRT development by creating an account on GitHub.
Fi just used official repo on windows today - latest auto1111
EHey community, question for you prompt specialists. Do any of you know what a "prompt modifier" is? I see it get used as a node in ComfyUI, and I see it get loaded in as a positive prompt. It has a "pre-prompt" and a "post-prompt" option. It's a nice addition. Is there a A1111 equivalent of it?
Thanks much!
Thanks much!
Dwhat does it do? change the prompt during generation? A1111 syntax is [elephant:mous:0.5]
EFrom what I can tell, yes, it prioritizes and changes the prompt during generation. When I did a metadata check, it looks like it's doing "pre-prompt - Text Input - Post prompt". But I'm not sure what logic it uses to prioritize.
Dwell the A1111 prompt can make elephats with mouse ears, if that's what you are looking for
ELet me see if I understand right. The "mous:0.5" is the weight of the prompt. what does the "elephant:mous" portion of the prompt do?
TWhat is that version/program of Comfy?
EGood question. Not sure. It's a workflow I saw online. I found the pre/post prompt concept inriguing.
EAlright, I did what I should have done 5 hours ago and asked Claude. There's a very clear explanation of what this is and it gave a great answer. I'll screenshot it here.
E
E
Fit looks to me nothing special :d
Fyou can do with styles
EIt’s not! It was just really bugging me that I didn’t know what the exact logic equivalent was in A1111 and my obsessive compulsive trait was eating at me. I tested things side by side after Claude’s answer. It checks out. I’m relaxed now!
Fwhat do you mean different?
Fyou mean logic? or you mean you cant do?
EI edited the message afterward I must have mentioned “different” accidentally. I meant I ran prompts side by side on comfy as well as A1111 after Claude’s answer. All checks out. It calmed my OCD. All is well on Beko land 
Thanks for checking in to make sure doc!
Thanks for checking in to make sure doc!
Fsubscribed envanto to add some more titles to videos :d
F
TI'm having difficulty making a dreambooth with another dataset, if someone can help/make it for me I'm willing to pay for it (apologies if I can't put this message here)
Dno this is (). the [a : b : 0.5] syntax above looks too similar, but does something completely different
Dit starts generation with a as the prompt, but after 50% of the generation steps, exchanges a with b