in some of your videos you showed previously some software that could generate prompts from an image
in some of your videos you showed previously some software that could generate prompts from an image, please help me to find it.
FAnthropic's most powerful model. Excels in complex tasks like coding, writing, analysis and visual processing. The context window has been shortened to optimize for speed and cost. For longer context messages, please try Claude-3.5-Sonnet-200k. The compute points value is subject to change.
Fit is free
Falso you can use joy captionm
SI think if was local software
Sthank you
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
F@slipkn67 this one amazing
D@Dr. Furkan Gözükara Dear, good morning, I hope you are doing well!
I need help, when I try to get the comparisons in the grid, I am realizing that it fails for some loras and it does not generate the image for me.
I need help, when I try to get the comparisons in the grid, I am realizing that it fails for some loras and it does not generate the image for me.
�I ran into issues before attempting to Generate Grid for 40 Loras with 1 prompt before. Was wanting to do the same exact thing you're doing so not completely sure yet if it's a SwarmUI or Vram issue
�I feel like 5 - 10 Loras at a time is the most Swarm can handle without issues, because I even tried to stack them on different axises but still didn't work. I don't know if it attempts to load all Loras on available Vram rather than load / unload one by one. So I gave up and used this Comfy workflow instead: https://civitai.com/models/652483/flux-lora-tester
Hello 
for whatever reason you want to show me appreciation, you can: 
Ko-Fi With this workflow you just save time. A lot of time. Have fun. To...
for whatever reason you want to show me appreciation, you can: Ko-Fi With this workflow you just save time. A lot of time. Have fun. To...
LDid anyone tried dreambooth Flux for a Style?
DI solved it by just leaving 2 GPUs instead of 4.
D
Donly 2
Fcan be corrupted lora
Fnot yet
Fi am testing onetrainer
Fgreat
Fok first 8 training onetrainer completed time to test
�For all 99 Loras? 
Z@Dr. Furkan Gözükara once the dataset has been prepared, it doesn't matter if we leave these fields blank does it?
Zcause the folder structure for img is 1_ohwx_man
Fyou dont need to use there
Ffirst results with one trainer almost identical kohya
Fstill evaluating
Ztesting with gguf right now with trained lora in kohya, results are decent so far
Zi only got to train up to 75 steps yesterday cause my son wanted to play beamng, doing a full training round today
Z
Fnice
Zi'm so happy they came up with the idea of quantizing these models, now everyone can have a chance to do it
Flooks like scotastic rounding of onetrainer yields better than kohya
Zdo you have a tutorial on that? cause i have no idea what that is
Fwill do
Fi am asking kohya whether we have or not
Zand will it run on low vram systems
Fnope they dont have fp8 yet
Fso you need massed compute
Zgot it
Fthey have nf4
Fwaiting images to be generated
Zoooo
Fok nf4 works too
Fnot too bad
DI have this problem, in grid usage, the prompt is separated by the commas, and I need it not to do so, but I don't want to remove the commas.
Dadd || <- solution
CHello all. I recently followed @Dr. Furkan Gözükara FLUX Lora tutorial. I got a very good result when testing the Lora with Flux dev model, but not so good when testing with Flux dev NF4 model. Is it something I'm doing wrong, or is it just not compatible?
Zyeah it doesn't work well with nf4, surprisingly tho it works well with the gguf dev models @Cypriot79
Fuse this