Software Engineering Courses (SECourses)•3y ago

I see you are checking Adafactor SDXL TE problem on Issue #855. Does this mean Adafactor is broken a

I see you are checking Adafactor SDXL TE problem on Issue #855. Does this mean Adafactor is broken at the moment, in light of TE? I thought it was beneficial to exclude SDXL Text Encoder from learning? Is Dadaptation a better option?

riccardo•10/6/23, 3:54 PM

can someone explain why google colab gives me this error:
ImportError Traceback (most recent call last)
<ipython-input-7-fdf29cbd2fa1> in <cell line: 22>()
20 import io
21 import math
---> 22 import timm
23 from IPython import display
24 import lpips

11 frames
/usr/local/lib/python3.10/dist-packages/PIL/ImageFont.py in <module>
32
33 from . import Image
---> 34 from ._util import isDirectory, isPath
35
36 LAYOUT_BASIC = 0

ImportError: cannot import name 'isDirectory' from 'PIL._util' (/usr/local/lib/python3.10/dist-packages/PIL/_util.py)

---------------------------------------------------------------------------
NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.

To view examples of installing some common dependencies, click the
"Open Examples" button below.

Jehex•10/6/23, 4:24 PM

Hello guy's, How can I run 2 instances of SD ( a1111 ) at the same time ?

Mmikemenders However, there is a striking difference between SD and RealisticVision, and betw...

EyeSpyBekoAI•10/6/23, 6:05 PM

Hey @mikemenders , try RV5 (not 5.1) with the 8400 VAE. I personally had better luck. And yes, SD1.5 vs. RV models are worlds apart.

EyeSpyBekoAI•10/6/23, 6:06 PM

Regardless though your pics look great!

Sandstormleader•10/6/23, 8:29 PM

so the --share command thing is broken i believe

Sandstormleader•10/6/23, 8:33 PM

Could not create share link. Please check your internet connection or our status page: https://status.gradio.app
Startup time: 13.1s (import torch: 1.2s, import gradio: 0.5s, import ldm: 0.3s, other imports: 0.8s, load scripts: 4.6s, create ui: 0.3s, gradio launch: 5.4s).

Mmikemenders https://github.com/Gourieff/sd-webui-reactor

EyeSpyBekoAI•10/6/23, 9:23 PM

Hey @mikemenders , this is interesting. Do you know the difference between this vs. adetailer, or have you tested them against each other?

JJehex Hello guy's, How can I run 2 instances of SD ( a1111 ) at the same time ?

thesilvermoth•10/6/23, 9:26 PM

I also want to know this. Do I need different SD folders?

SSandstormleader so the --share command thing is broken i believe

thesilvermoth•10/6/23, 9:27 PM

It will only work if SD is on local. You can't use that command on RunPod or Kaggle

Sandstormleader•10/6/23, 9:36 PM

I only use local

Sandstormleader•10/6/23, 9:37 PM

Do I need an extension or something as well?

PIXLE•10/6/23, 9:45 PM

Question for any of you: I have trained LORAs for my wife, myself, and my two kids. Whats the best way to compose a family portrait or group photo? Any good tutorials out there? I haven't found much.

Sandstormleader•10/6/23, 9:48 PM

i thought about doing something lie that but my approach was going to train multiple concepts within the same model with dreambooth

PPIXLE Question for any of you: I have trained LORAs for my wife, myself, and my two ki...

EyeSpyBekoAI•10/6/23, 9:54 PM

@xPIXELPAPIx you'll need to use the Loras in a single prompt. So for example it'll be something like "husband:lora" sitting next to "wife:lora" with "son:lora" chasing "daughter:lora" in a living room. Something along those lines. I'm not too experienced with complicated Loras myself but it should be something along those lines and there are a few youtube videos out there of people using multiple loras.

Kking Hi @here

Furkan Gözükara SECourses•10/6/23, 10:10 PM

SSandstormleader @Dr. Furkan Gözükara have you tested results of data photos having a beard vs no...

Furkan Gözükara SECourses•10/6/23, 10:10 PM

i dont think you need

Furkan Gözükara SECourses•10/6/23, 10:11 PM

i can get shaved photo :d

Rrafraf Hey @Dr. Furkan Gözükara got amazing results for 12 image character LoRA using A...

Furkan Gözükara SECourses•10/6/23, 10:16 PM

i will explore lora adafactor later. currently finding best hyper params for dreambooth.

Rrafraf I see you are checking Adafactor SDXL TE problem on Issue #855. Does this mean A...

Furkan Gözükara SECourses•10/6/23, 10:17 PM

i got super interesting finding with TE still researching. but we really need different LR for TE : https://twitter.com/GozukaraFurkan/status/1710416135747748150

Furkan Gözükara (@GozukaraFurkan)

SDXL text encoder turns out to be very strong and becomes so quickly overtrained

Twitter

•

10/6/23, 10:05 PM

Rriccardo can someone explain why google colab gives me this error: ImportError ...

Furkan Gözükara SECourses•10/6/23, 10:18 PM

which colab?

JJehex Hello guy's, How can I run 2 instances of SD ( a1111 ) at the same time ?

Furkan Gözükara SECourses•10/6/23, 10:18 PM

yes

FFurkan Gözükara SECourses i got super interesting finding with TE still researching. but we really need di...

rafrafOP•10/6/23, 10:18 PM

yea, i saw the thread. Sounds like very difficult to test for TE? Have you got some idea about it? I thought there being two TE's made it impossible practically

Furkan Gözükara SECourses•10/6/23, 10:18 PM

but it really doesnt worth unless you have 2 gpu

Rrafraf yea, i saw the thread. Sounds like very difficult to test for TE? Have you got s...

Furkan Gözükara SECourses•10/6/23, 10:18 PM

i dont know still searching

SSandstormleader Could not create share link. Please check your internet connection or our status...

Furkan Gözükara SECourses•10/6/23, 10:19 PM

ye it wasnt working yesterday

Furkan Gözükara SECourses•10/6/23, 10:19 PM

and today for a while

rafrafOP•10/6/23, 10:19 PM

I am testing AdamW8bit quality with batch sizes 1, 2, 4, 6. I am seeing what is the best sweetspot for quality vs speed of training

PPIXLE Question for any of you: I have trained LORAs for my wife, myself, and my two ki...

Furkan Gözükara SECourses•10/6/23, 10:19 PM

you can try latent couple extension

Rrafraf I am testing AdamW8bit quality with batch sizes 1, 2, 4, 6. I am seeing what is ...

Furkan Gözükara SECourses•10/6/23, 10:19 PM

yep

Furkan Gözükara SECourses•10/6/23, 10:19 PM

this is necessary for companies

Furkan Gözükara SECourses•10/6/23, 10:19 PM

i will also research this for a company

SSandstormleader @Dr. Furkan Gözükara have you tested results of data photos having a beard vs no...

Furkan Gözükara SECourses•10/6/23, 10:23 PM

it adds beard hard

i tested not easy to get without a beard

rafrafOP•10/6/23, 10:25 PM

i got pretty impressive results with batch 2, which drove SDXL adamW8bit training by by 30% and bs 4 by 60%. For me it's somewhere in there. Need to check flexibility, but i am sure there are other experimental variables to test. Here's dataset of 12 images, vs xy plot with batch 2 up to 9 epochs (5000 steps), 5 hours vs 9 hours batch size 1

xyz_grid-0007-20231006232009-__lora_4du7_0103-lora_adamw8bit_40rep2bs-dim128-alpha64-000001_0.8___closeup_portrait_of_4du7_woman_green_studio_background_st.jpg

rafrafOP•10/6/23, 10:26 PM

interesting the level of variance in the dataset gets picked up in dim 128/alpha64. I guess the ultimate test will come when i run this LoRA on a Art Style model

REEM•10/6/23, 10:28 PM

@Dr. Furkan Gözükara Hello, I'm trying to train LoRa on objects but it's not working. What's the reason? All the sources I saw were for faces

RREEM @Dr. Furkan Gözükara Hello, I'm trying to train LoRa on objects but it's not wor...

Furkan Gözükara SECourses•10/6/23, 10:30 PM

well it works but object training is a different methodology

Furkan Gözükara SECourses•10/6/23, 10:30 PM

i can give private consultation

FFurkan Gözükara SECourses well it works but object training is a different methodology

REEM•10/6/23, 10:33 PM

How?

FFurkan Gözükara SECourses hi

king•10/6/23, 10:36 PM

I am looking for any dev task.
Cold you help me?

Kking I am looking for any dev task. Cold you help me?

Furkan Gözükara SECourses•10/6/23, 10:37 PM

sorry i dont have any tasks

PIXLE•10/6/23, 10:40 PM

@Dr. Furkan Gözükara If I am training a model of my son, should I use reference images of boys his age instead of the ones I would use to train my own LORA?

Sandstormleader•10/6/23, 10:48 PM

Sandstormleader•10/6/23, 10:48 PM

looking better now.

Furkan Gözükara SECourses•10/6/23, 10:53 PM

https://twitter.com/GozukaraFurkan/status/1710427890070786480

PPIXLE @Dr. Furkan Gözükara If I am training a model of my son, should I use reference ...

Furkan Gözükara SECourses•10/6/23, 10:53 PM

if you can find yes better class dataset

Furkan Gözükara SECourses•10/6/23, 10:53 PM

BS 2 brings 82% speed increase on A100

Sandstormleader•10/6/23, 11:26 PM

EEyeSpyBekoAI Hey @mikemenders , try RV5 (not 5.1) with the 8400 VAE. I personally had better ...

mikemenders•10/6/23, 11:43 PM

Thank you for tip, I will try it soon!

EEyeSpyBekoAI Hey @mikemenders , this is interesting. Do you know the difference between this ...

mikemenders•10/6/23, 11:44 PM

I've had no luck with adetailer, but ReActor works for me, and if the shape of the face fits, it can be a really good tool for creating new images to go with existing ones.

Mmikemenders I've had no luck with adetailer, but ReActor works for me, and if the shape of t...

EyeSpyBekoAI•10/6/23, 11:47 PM

Interesting. Thanks for the heads up I'll give it a shot.

I see you are checking Adafactor SDXL TE problem on Issue #855. Does this mean Adafactor is broken a

Similar Threads