Software Engineering Courses (SECourses)•7mo ago

Dr Furkan, I saw you released an update for SDXL finetunning today, 20/june. Are you planning any up

Dr Furkan, I saw you released an update for SDXL finetunning today, 20/june. Are you planning any update for FLUX finetunning or your last release for FLUX are the best yet?

Ppianogospel Dr Furkan, I saw you released an update for SDXL finetunning today, 20/june. Are...

Furkan Gözükara SECourses•6/20/25, 11:53 AM

i recently tested flux fine tuning with more than 20 experiments

Furkan Gözükara SECourses•6/20/25, 11:53 AM

none yielded better

pianogospelOP•6/20/25, 1:38 PM

Dr using your tutorial and trainning a FLUX finetunning in Massed Compute, about choosing the GPU, is there a better GPU to choose? For better I'm only considering faster trainning and the relation cost x benefit, I know that the prices increase. I don't understand one thing, if I choose a H100 x8, or a 8 H100 boards, is it eight times faster than an 1 H100? or it doesn't work this way? Can you tell me what is the best board option in Massed Compute, ignoring price, for a 100 images finetunning of a person?

Ppianogospel Dr using your tutorial and trainning a FLUX finetunning in Massed Compute, about...

Furkan Gözükara SECourses•6/20/25, 1:51 PM

for lora it will be almost 8 x

Furkan Gözükara SECourses•6/20/25, 1:51 PM

but for dreambooth last time i tested it wasnt working properly

Furkan Gözükara SECourses•6/20/25, 1:51 PM

best price performance is RTX a6000 ADA i think

pianogospelOP•6/20/25, 2:33 PM

In your post, about Dreambooth finetunning, you said: "...single RTX 4090 is almost same speed as RTX A6000 for Fine Tuning." Does it mean that if I have a RTX 4090 can I train, locally, at the same speed as a RTX a6000 that has 48gb of vram, double vram than mine 4090?

Ppianogospel In your post, about Dreambooth finetunning, you said: "...single RTX 4090 is alm...

Furkan Gözükara SECourses•6/20/25, 2:42 PM

4090 is faster when fits into VRAM. current not fitting but with block swapping it matches almost

Furkan Gözükara SECourses•6/20/25, 2:42 PM

what is your second / it ?

Furkan Gözükara SECourses•6/20/25, 2:42 PM

5090 is almost twice faster than A6000 since fits into VRAM fully

pianogospelOP•6/20/25, 2:56 PM

So, if I choose the file 16GB_GPU_15150MB_11.6_second_it_Tier_1 to train a FLUX finetunning, my 4090 is better than a RTX A6000? Is it right? I imagine that I can't choose the 24GB_GPU_22900MB_7.3_second_it_Tier_1 because it's to close to use the full amount of memory, right?

Ppianogospel So, if I choose the file 16GB_GPU_15150MB_11.6_second_it_Tier_1 to train a FLUX ...

Furkan Gözükara SECourses•6/20/25, 5:09 PM

true

pianogospelOP•6/21/25, 3:20 PM

Dr, is 6.36s/it a good velocity to 16GB_GPU_15150MB_11.6_second_it_Tier_1 config in a 4090, locally?

Ppianogospel Dr, is 6.36s/it a good velocity to 16GB_GPU_15150MB_11.6_second_it_Tier_1 conf...

Furkan Gözükara SECourses•6/21/25, 6:09 PM

yes really good

PabloMoises•6/23/25, 11:48 PM

Hi! Does anyone know if it's possible to upload a fine-tuned model to Replicate to use it via API? And if someone knows, could you give me a mini step-by-step tutorial on how to do it?

Thanks a lot!

PPabloMoises Hi! Does anyone know if it's possible to upload a fine-tuned model to Replicate ...

Furkan Gözükara SECourses•6/24/25, 10:20 PM

yes possible

Furkan Gözükara SECourses•6/24/25, 10:20 PM

but you aint gonna find tutorial it is only useful for companies

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•6/25/25, 8:59 AM

if i were to train a finetune on say norwegian women, to better capture their looks, say i have 5 models with maybe 10 photos each i use for this. would this be an example of where we use captions? @Furkan Gözükara SECourses

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭if i were to train a finetune on say norwegian women, to better capture their lo...

Furkan Gözükara SECourses•6/25/25, 10:57 AM

yes for general fine tuning you can use captions

Furkan Gözükara SECourses•6/25/25, 10:57 AM

but you need more photos

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•6/25/25, 11:03 AM

Ok. The other plan was to train 5 woman loras, then use all on low weight, without triggerword and prompt the womans looks.

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭Ok. The other plan was to train 5 woman loras, then use all on low weight, with...

Furkan Gözükara SECourses•6/25/25, 11:26 AM

each lora may overwrite each other but you can try

FFurkan Gözükara SECourses yes possible

PabloMoises•6/25/25, 4:23 PM

Thank you very much! I'm going to try anyway because I really need it, if anyone knows of something that could help me, I would truly appreciate it.

Flame火•6/27/25, 1:09 PM

Hey everyone, looking for a quick sanity check on my training parameters before I commit my resources. I've done my best to prep everything properly.

Model: Flux (via fluxgym)

Goal: A flexible character LoRA with two distinct styles (style_A and style_B).

Dataset: 98 high-quality, curated images.

Variant A (Core/Black Hair): ~55 images

Variant B (Secondary/Lighter Hair): ~43 images

Shot Distribution: ~14% full-body, ~38% medium, ~48% close-ups.

My Proposed Parameters: Repeats: 4, Epochs: 24, Batch Size: 2, Total Steps: ~5040 (approx. 51 steps per image), Unet LR: 1e-4, Text Encoder LR: 5e-5, Scheduler: cosine, Save every N epochs: 2

Does this recipe look solid for a high-quality Flux character LoRA? Any red flags or suggestions? Thanks!

FFlame火 Hey everyone, looking for a quick sanity check on my training parameters before ...

Furkan Gözükara SECourses•6/27/25, 1:47 PM

well i only can recommend our researched training config

Furkan Gözükara SECourses•6/27/25, 1:48 PM

but image distribution is sounding good

Furkan Gözükara SECourses•6/27/25, 1:48 PM

however different styles will get bleed into each other

FFurkan Gözükara SECourses well i only can recommend our researched training config

Flame火•6/27/25, 2:04 PM

Thanks, I really appreciate you taking the time.

Your point about style bleeding is exactly why I set up the style_A and style_B tags in my captions—I'm hoping that gives me the control to call them separately.

you mentioned your researched training config. Is that publicly available anywhere? I'd love to read up on it and compare it with my setup to make sure I'm on the right track.

FFlame火 Thanks, I really appreciate you taking the time. Your point about style bleedi...

Furkan Gözükara SECourses•6/27/25, 2:06 PM

with flux they bleed sadly

Furkan Gözükara SECourses•6/27/25, 2:06 PM

my configs are on patreon

Furkan Gözükara SECourses•6/27/25, 2:06 PM

https://youtu.be/FvpWy1x5etM

YouTubeSECourses

FLUX Full Fine-Tuning / DreamBooth Training Master Tutorial for Win...

If you want to train FLUX with maximum possible quality, this is the tutorial looking for. In this comprehensive tutorial, you will learn how to install Kohya GUI and use it to fully Fine-Tune / DreamBooth FLUX model. After that how to use SwarmUI to compare generated checkpoints / models and find the very best one to generate most amazing image...

FFurkan Gözükara SECourses with flux they bleed sadly

Flame火•6/27/25, 2:38 PM

Based on your warning, I'm now planning to train two separate loras as the safest path

Before I separate the dataset, I had one last quick question about the "style bleed." Was your experience with this happening on Flux even when the underlying facial identity was identical across both styles?

I ask because my dataset uses the exact same face model for both my "styles," with the only changes being things like hair color and subtle makeup. I'm trying to understand if flux is so powerful that it still blends even those subtle, correlated changes, or if the bleeding happens mostly when training on two more distinct concepts (like different characters or art styles).

FFlame火 Based on your warning, I'm now planning to train two separate loras as the safes...

Furkan Gözükara SECourses•6/27/25, 9:46 PM

yes with flux if what you train has same genre it will bleed

Furkan Gözükara SECourses•6/27/25, 9:46 PM

like 2 human

Furkan Gözükara SECourses•6/27/25, 9:46 PM

like 3d drawing style

Furkan Gözükara SECourses•6/27/25, 9:46 PM

you can train a dog and a human together though

pianogospelOP•6/28/25, 2:16 PM

Hi Dr. I want to try a full finetune / dreambooth with Flux SCHNELL. Do you have configs to use in Kohya SS for this model or the configs for Flux DEV works in the same way?

gio3000.•6/28/25, 3:20 PM

Hi, can I finetune flux kontext with dreamboth ?

Ppianogospel Hi Dr. I want to try a full finetune / dreambooth with Flux SCHNELL. Do you hav...

Furkan Gözükara SECourses•6/28/25, 9:42 PM

our current configs works

Ggio3000.Hi, can I finetune flux kontext with dreamboth ?

Furkan Gözükara SECourses•6/28/25, 9:42 PM

not with kohya yet

Furkan Gözükara SECourses•6/28/25, 9:42 PM

also it reuqires special dataset

Kat-e•6/29/25, 12:15 AM

Anyone know what would cause my training to just stop randomly? This has happened numerous times since I upgraded to a 5090 a few weeks back. No errors are displayed and there seems to be nothing in particular triggering the event. I reinstalled everything and it runs smoothly...until it just stops and gpu utilization drops to zero. Once this occurs I'm unable to create a backup and have to restart onetrainer and resume from backup. Very frustrating because I have to keep restarting the training and can't leave and let it cook

KKat-e Anyone know what would cause my training to just stop randomly? This has happene...

Furkan Gözükara SECourses•6/29/25, 12:36 AM

yes out of RAM

Furkan Gözükara SECourses•6/29/25, 12:36 AM

it can cause crash

Kat-e•6/29/25, 12:40 AM

I just restarted a training and RAM (while it's working is only at 22 (I have 64). VRAM however is at 31.5

KKat-e I just restarted a training and RAM (while it's working is only at 22 (I have 64...

Furkan Gözükara SECourses•6/29/25, 12:50 AM

how much watt you using?

Furkan Gözükara SECourses•6/29/25, 12:50 AM

by the way normally it uses around 28.5gb vram

Furkan Gözükara SECourses•6/29/25, 12:50 AM

if all 1024x1024px

Kat-e•6/29/25, 12:59 AM

Watt- fluctuating between 350-500 approximately if I'm looking at it right. I'm doing an SDXL training btw. All 1024x1024, batch size 2, accumulation steps 3 just experimenting. But ever since I installed 5090 Onetrainer just randomly stops responding.

KKat-e Watt- fluctuating between 350-500 approximately if I'm looking at it right. I'm ...

Furkan Gözükara SECourses•6/29/25, 1:02 AM

ah ye that is why

Dr Furkan, I saw you released an update for SDXL finetunning today, 20/june. Are you planning any up

Similar Threads