it has only 6 gb vram
it has only 6 gb vram
N: ERROR Error while calling W&B API: user is not logged in (<Response [401]>)
Problem at: /workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/tracking.py 294 init
wandb: ERROR The API key is either invalid or missing, or the host is incorrect. To resolve this issue, you may try running the 'wandb login --host [hostname]' command. The host defaults to 'https://api.wandb.ai' if not specified. (Error 401: Unauthorized)
Traceback (most recent call last):
File "/workspace/kohya_ss/./train_db.py", line 481, in <module>
train(args)
File "/workspace/kohya_ss/./train_db.py", line 258, in train
accelerator.init_trackers("dreambooth" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/accelerator.py", line 564, in _inner
return PartialState().on_main_process(function)(*args, kwargs)
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/accelerator.py", line 2096, in init_trackers
self.trackers.append(tracker_init(project_name, init_kwargs.get(str(tracker), {})))
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/tracking.py", line 85, in execute_on_main_process
return function(self, *args, kwargs)
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/tracking.py", line 294, in init
self.run = wandb.init(project=self.run_name, kwargs)
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 1164, in init
raise e
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 1145, in init
run = wi.init()
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 764, in init
raise error
wandb.errors.AuthenticationError: The API key is either invalid or missing, or the host is incorrect. To resolve this issue, you may try running the 'wandb login --host [hostname]' command. The host defaults to 'https://api.wandb.ai' if not specified. (Error 401: Unauthorized)
NI recently tried to start training in Kohya and I got that error, do you know what it could be?
Sis there anyway for me to run a seperate instance of automatic1111 so that i can have the old version with working dreambooth as well as a updated version @Dr. Furkan Gözükara
Fit says wan db api error
Finstall into a new folder
Fthats it
Sthank you, do you have a link to the latest installer?
Saslo did you ever find out if it was possble to run from an external hdd?
Fyes external hdd working
FFauto installer windows
SPerfect thank you
Myes of course, I mean in your video its was Dreambooth and Dreambooth lora ( exact name) maybe renamed to lora only in the newest version .
Fthey renamed
MI have now tried the Prodigy Lora training with another parameter that is also recommended for SDXL. In general, in both cases, epoch 10 (the last one) gives (for me) absolutely satisfactory and stable results.
- This is the optimizer args that I am using now: "use_bias_correction=True" "weight_decay=0.01" "decouple=True" "betas=(0.9,0.999)" "d_coef=2" "safeguard_warmup=False"
It makes the model more flexible and works well on Clip Skip 1. The way I look at it, 3-6 epochs already give good results, at higher epochs you can see one of the characteristics of overdrawing, the freckling.
- this is what I used yesterday: "use_bias_correction=True" "weight_decay=0.5" "decouple=True" "betas=(0.9, 0.99)" "d_coef=2" "safeguard_warmup=False" (here the weight_decay value is stronger)
This is also very good from 3, and even 10 didn't burn. But this gives an absolutely strong, realistic model even at 512x512 resolution, without Hires Fix. This is worth training on Clip Skip 2 to make it flexible. I got such a perfect result that it not only restored the tattoo to its position, but even its shape. However, it does not like getting naked because of this.
So if you want very realistic images, the latter is better, if you want more flexible, the former.
Update1: Epoch 4 seems to be the most flexible out of the 10, after that you need stronger prompts if you want to see not only the face but also the body.
- This is the optimizer args that I am using now: "use_bias_correction=True" "weight_decay=0.01" "decouple=True" "betas=(0.9,0.999)" "d_coef=2" "safeguard_warmup=False"
It makes the model more flexible and works well on Clip Skip 1. The way I look at it, 3-6 epochs already give good results, at higher epochs you can see one of the characteristics of overdrawing, the freckling.
- this is what I used yesterday: "use_bias_correction=True" "weight_decay=0.5" "decouple=True" "betas=(0.9, 0.99)" "d_coef=2" "safeguard_warmup=False" (here the weight_decay value is stronger)
This is also very good from 3, and even 10 didn't burn. But this gives an absolutely strong, realistic model even at 512x512 resolution, without Hires Fix. This is worth training on Clip Skip 2 to make it flexible. I got such a perfect result that it not only restored the tattoo to its position, but even its shape. However, it does not like getting naked because of this.
So if you want very realistic images, the latter is better, if you want more flexible, the former.
Update1: Epoch 4 seems to be the most flexible out of the 10, after that you need stronger prompts if you want to see not only the face but also the body.
Fjust found out amazing thing
Fwarp of cloudflare
Ffixes pip install network issues haha
Ffree vpn
EAnyone have any idea why runpodctl slows down when it arrived at 95% completion? i get 80mb/sec at start, 14mb/sec after 20% and its gradually slowing down as time progresses down to 100kb/sec
Fye it is really slow
Fnow i prefer hugging face
EI got this bad resaults in dreambooth training, did 8000 steps, 24 images, 42 repeats, 8 epochs
Why did the face come out liek this ? (happens in all generations)
Why did the face come out liek this ? (happens in all generations)
Seven on closer up renders?
Eim using Adetailer for face, but ill try close up
Eyes even on closeups
Swhat is your cfg at?
E7
Stry 3 or 4
Si never used auto detailer so not sure on that but sometimes cfg needs to be lower
Fdid you try previous epochs?
Fi am working on an auto captioner best one
Eyes, i saved every 2nd epoch so i have 2, 4, 6, 8
but all of them makes this kinds of weird bugs
but all of them makes this kinds of weird bugs
FViT-bigG-14', 'laion2b_s39b_b160k
Fdm me your print training command
EI get this two photos when i try "photo of ohwx man" cfg 5 no adtailer
Fsomething wrong with your training possibly
Esent you the command on dm
F1 more thing
Fcan you test with
FDPM++ 2M SDE Karras
F40 steps
AI have 8 gb VRAM:)
And this settings really helped for me
And this settings really helped for me
Fnice
Fyes --medvram super important
