Software Engineering Courses (SECourses)•15mo ago

Tomorrow I will ask you guys to help me choose best checkpoint as , I cant decide lol. Then I will t

Tomorrow I will ask you guys to help me choose best checkpoint as , I cant decide lol.
Then I will try reduce size from 23 to 11.

Ccyberbol Tomorrow I will ask you guys to help me choose best checkpoint as , I cant decid...

Furkan Gözükara SECourses•10/24/24, 9:38 PM

ye it is hard

Furkan Gözükara SECourses•10/24/24, 9:38 PM

with flux all images are good

Furkan Gözükara SECourses•10/24/24, 9:38 PM

size reduce works 0 quality loss

cyberbolOP•10/24/24, 9:39 PM

Thats way I want reduce. I managed already send all checkpoints to HugginsFace. 10x23gb, so when I decide which one I like... will download to local PC and change size. Not watchen yet video about reducing size , so I hope its possible on local PC, or must be done at Massed Comp

Ccyberbol Thats way I want reduce. I managed already send all checkpoints to HugginsFace. ...

Furkan Gözükara SECourses•10/24/24, 9:40 PM

totally possible

cyberbolOP•10/24/24, 9:42 PM

ok, I think I need cancel genrate GRID of Stylized as its still almost 2H and Im very sleepy. lol. If I copy from Massed Comp folder GRID, can I continue tomorrow generate this GRID ?

Ccyberbol ok, I think I need cancel genrate GRID of Stylized as its still almost 2H and Im...

Furkan Gözükara SECourses•10/24/24, 9:48 PM

i think should work

Furkan Gözükara SECourses•10/24/24, 9:48 PM

make sure to backup everything on massed compute gets deleted

Furkan Gözükara SECourses•10/24/24, 9:48 PM

once terminate

cyberbolOP•10/24/24, 9:49 PM

i know, checked 10x times lol If definitelly all my 23gb files are safe on HuginsFace haha

cyberbolOP•10/24/24, 9:54 PM

Goodnight,
Dobranoc,
İyi geceler

Zet•10/24/24, 9:57 PM

Dobranoc

masood•10/25/24, 6:38 AM

guys do we have some prompts that works really good with generating images using stable diffusion models ?

Mmasood guys do we have some prompts that works really good with generating images using...

Furkan Gözükara SECourses•10/25/24, 8:58 AM

i sometimes check civitai

Furkan Gözükara SECourses•10/25/24, 8:58 AM

to see prompts

Furkan Gözükara SECourses•10/25/24, 8:58 AM

also after writing some good working prompts

Furkan Gözükara SECourses•10/25/24, 8:58 AM

using chatgpt to get new ones works

masood•10/25/24, 10:06 AM

thanks

masood•10/25/24, 10:06 AM

sorry guys i am asking alot, but i need to.

masood•10/25/24, 10:07 AM

anyone have got some good flux-dev-de-destill OneTrainer finetuning configurations ?

theJindesan•10/25/24, 3:10 PM

@Furkan Gözükara SECourses Could you check v9 configs for 8GB again.. with Torch 2.5, i get out of memory errors, With 2.4 - it works but very slow. It was working on an earlier version of your 8GB configs. File "E:\StabilityMatrix-RAID\Kohya_Flux\kohya_ss\sd-scripts\library\flux_models.py", line 830, in _forward
attn = attention(q, k, v, pe=pe, attn_mask=attn_mask)
File "E:\StabilityMatrix-RAID\Kohya_Flux\kohya_ss\sd-scripts\library\flux_models.py", line 449, in attention
x = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=attn_mask)
torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.90 GiB. GPU 0 has a total capacity of 8.00 GiB of which 0 bytes is free. Of the allocated memory 8.05 GiB is allocated by PyTorch, and 2.07 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)
steps: 0%|

TtheJindesan <@205854764540362752> Could you check v9 configs for 8GB again.. with Torch 2.5,...

Furkan Gözükara SECourses•10/25/24, 8:45 PM

can you try with SDPA?

Furkan Gözükara SECourses•10/25/24, 8:47 PM

i am testing on linux now to test

Furkan Gözükara SECourses•10/25/24, 8:47 PM

exact vram

Furkan Gözükara SECourses•10/25/24, 8:47 PM

you have shared vram enabled right?

Furkan Gözükara SECourses•10/25/24, 8:48 PM

ye it still uses 7gb vram

FFurkan Gözükara SECourses can you try with SDPA?

theJindesan•10/25/24, 9:39 PM

I think it has to do with Torch 2.5? Does it work on windows? Shared VRAM should be enabled, system has 64GB, and Task Manager shows 40GB for GPU Memory (8+32), shared GPU memory shows 32GB

TtheJindesan I think it has to do with Torch 2.5? Does it work on windows? Shared VRAM shoul...

Furkan Gözükara SECourses•10/25/24, 9:41 PM

yes it works

Furkan Gözükara SECourses•10/25/24, 9:41 PM

torch 2.5 and xformers is now official

FFurkan Gözükara SECourses torch 2.5 and xformers is now official

theJindesan•10/25/24, 9:42 PM

What else can i try? It worked before, but not after using latest torch and configs.. clean SS install using your scripts

TtheJindesan What else can i try? It worked before, but not after using latest torch and con...

Furkan Gözükara SECourses•10/25/24, 9:42 PM

ye clean and let me know

Furkan Gözükara SECourses•10/25/24, 9:42 PM

i will also now test on my RTX 3060

FFurkan Gözükara SECourses ye clean and let me know

theJindesan•10/25/24, 9:42 PM

I did try Clean.. This is a clean install.

TtheJindesan I did try Clean.. This is a clean install.

Furkan Gözükara SECourses•10/25/24, 9:42 PM

did you try 6 gb config?

theJindesan•10/25/24, 9:43 PM

yes. also gives out of memory

Furkan Gözükara SECourses•10/25/24, 9:43 PM

so something wrong in your case

Furkan Gözükara SECourses•10/25/24, 9:43 PM

the only changed config is SPDA to xformers

theJindesan•10/25/24, 9:44 PM

with SPDA setting.. RuntimeError: CUDA error: out of memory
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with TORCH_USE_CUDA_DSATORCH_USE_CUDA_DSA to enable device-side assertions.

TtheJindesan with SPDA setting.. RuntimeError: CUDA error: out of memory CUDA kernel errors m...

Furkan Gözükara SECourses•10/25/24, 9:44 PM

your computer is the issue in that case

Furkan Gözükara SECourses•10/25/24, 9:44 PM

because we use torch 2.5 for a month like now

Furkan Gözükara SECourses•10/25/24, 9:45 PM

i am doing a fresh install

Furkan Gözükara SECourses•10/25/24, 9:45 PM

normally you shouldnt get OOM

Furkan Gözükara SECourses•10/25/24, 9:45 PM

when shared VRAM enabled

theJindesan•10/25/24, 9:45 PM