so tier 1 ok thanks
so tier 1 ok thanks
Fdefinitely
Fyes
Lrepeats the same? 150-200 or i need to lower?
Abtw doctor for multitalk very good and stable results gives the default t2v lighxv2 rank 32
Flower
Ffor 300 images do like maximum 100 epoch and compare
Fwell i didnt test to many times
Aim closing my h100 soon
Aget ready to test
Athanks
Fso your h100 failing at sage attention?
For what error you getting
Ano h100 good but c10 error
Aon first launch
A
U@Furkan Gözükara SECourses Hello,
I was wondering if you have a python code that I can refer/modify for finetuning instead of using the kohya gui? I want to load the parameters and train without having to open the interface and load the parameters all the time.
I was wondering if you have a python code that I can refer/modify for finetuning instead of using the kohya gui? I want to load the parameters and train without having to open the interface and load the parameters all the time.
Fgui runs cmd code literally
Fpay attention to cmd
UYeah, I tried it, but the results I get vary. However, from the GUI, I get good results. I’ll check if the code uses the correct parameters again.
Fye kohya is based on cmd command
Fso you need to make it accurate
UI was able to get it working.
Also kohya does not support chroma finetuning do you know any other open source model that supports chroma finetuning?
Also kohya does not support chroma finetuning do you know any other open source model that supports chroma finetuning?
FGitHub
The ultimate training toolkit for finetuning diffusion models - ostris/ai-toolkit
Fi am not sure though
Falso make a reply here
FGitHub
With the FLUX.1-schnell based Chroma model gaining popularity is there any chance of it being added, for lora trainng ?
UI tried ai tool kit but I could only find lora training. I will check again.
U"I’d like to support this request. It would be beneficial if the developers could add support for the Chroma model.
Chroma is an 8.9B parameter model based on FLUX.1-schnell. More information can be found here: huggingface.co/lodestones/Chroma
Additionally, ai-toolkit and the diffusion-pipe project at github.com/tdrussell/diffusion-pipe already supports Chroma. (LoRA/fine-tuning).
Thank you for the fantastic sd-scripts project; your work is greatly appreciated!"
I came across this comment by blackmagix24. I will give diffusion pipe a shot. if you happen to test it in the future, I would love to see a video from you for chroma fine tuning along with a comparison with flux dev
Chroma is an 8.9B parameter model based on FLUX.1-schnell. More information can be found here: huggingface.co/lodestones/Chroma
Additionally, ai-toolkit and the diffusion-pipe project at github.com/tdrussell/diffusion-pipe already supports Chroma. (LoRA/fine-tuning).
Thank you for the fantastic sd-scripts project; your work is greatly appreciated!"
I came across this comment by blackmagix24. I will give diffusion pipe a shot. if you happen to test it in the future, I would love to see a video from you for chroma fine tuning along with a comparison with flux dev
Fit is just weaker
Fwhat makes flux different is that it is literally built to make money
Fso it is really good
Uyeah, when I tried chroma the images weren't as good as flux dev, but on the bright side chroma has apache license and I was wondering if finetuned well, will the results be in par with flux dev.
Fkohya adding it
FGitHub
Thank you lodestones for the high quality model.
--model_type option is added to flux_train_network.py. For Chroma, --model_type chroma is required.
According to lodestones, --apply_t5_attn_mask --...
--model_type option is added to flux_train_network.py. For Chroma, --model_type chroma is required.
According to lodestones, --apply_t5_attn_mask --...
LMister Doctor, if i am to train on another SDXL model with baked vae, should i delete additional vae from preset? Tier1_48_GB_Faster_v2
this one
VAE (Optional: Path to checkpoint of vae for training)
stabilityai/sdxl-vae
this one
VAE (Optional: Path to checkpoint of vae for training)
stabilityai/sdxl-vae
Lthank you
and if i do batch size=2, do i need to adjust learning rates? like double them
and if i do batch size=2, do i need to adjust learning rates? like double them
Ftry square root of batch size
MHi! I just downloaded Kohya and all necessary models in my Runpod. How can I use best configs json file(e.g. 48GB_GPU_Quality_Tier_1_40000MB_9.3_Second_IT_Trains_T5_and_T5_Attention) in Kohya?
Should I manually put all the parameters?
Should I manually put all the parameters?
MAnd I tried this config with L40S, but keep shows "torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 72.00 MiB.
GPU 0 has a total capacity of 44.52 GiB of which 24.25 MiB is free.
Process 165082 has 44.49 GiB memory in use."
GPU 0 has a total capacity of 44.52 GiB of which 24.25 MiB is free.
Process 165082 has 44.49 GiB memory in use."
Fyes use lesser config
Fprobably your images are not all 1024x1024
MI just checked and it's all 1024 * 1024. Is there any possibility that I'm facing OOM even though I'm using L40S(And I checked the memory usage right before the training)
Fwell it uses 40 GB so you shouldnt
Fcan you try restart machine and try again?
Fbut maybe kohya made some updates and now it uses more VRAM
MI'll try and let you know!
M@Furkan Gözükara SECourses When I change config to "cache_latents = false / cache_latents_to_disk = false", GPU usage dropped to 36GB. Is there any quality drop if I change it like this?
