A simple and quick lora trainer setup, I got 1it/s after 5min of training on a 512px training, batch 1, so it's pretty fast.
You can use flux1_devFP8Kijai11GB.safetensors as well as the regular flux1-dev.safetensors model,
flux1-dev is slightly faster but take abit more time to load, the VRAM usage is the same, around 16.3 GB of VRAM, set the HighVram to false doesn't change anything on this side.
Set the "Split Mode" to "true," will use a maxium of 41% of 24GB, so around 10GB of VRAM usage,
Use it only If you have less than 16GB, If you have 16GB you should try "false" before, and see if you got an OOM. Setting this to "true" and the training will last 110 minutes instead of 20 on this particular example.
I made 3 custom nodes that you will find on comfyUI manager (missing nodes)
To train on specific single blocks, separated by Layer1 and Layer2
Added validation previews and loss, and save the preview and training arguments into the destination folder
600 steps at 512 resolution, took me 15 min to train
Description
FAQ
Comments (14)
How do you set the images up for training? Is it like kohya_ss or a different format? Can you give us some, like, information here? Because it looks interesting and I'd like to try it. So, yeah. thanks my the way...
Works perfectly with small GPU (3060 12GB rtx). As of 17/09/2024 all nodes were find easily and downloaded with comfyui manager. And i just trained a lora yesterday at night, it takes me 3 and half to 4 hours of training with a data set of approximately 80 pictures. The quality is the same as a Flux lora trained on Civitai. It's increddible that i can use the same kind of 512*512 dataset that i use for SD.1.5 😀.
Hello, is it possible to use this Trainer Node in SwarmUI? If it is possible, could you describe step by step the installation?
where do you set the epochs at?
[FIXED, simply fresh ComfyUI portable install solved the issue!] couldn't run because if this error (I'm probably missing some files)
module 'torch' has no attribute 'float8_e4m3fnuz'
just in case someone else had same issue, that's how you fix it :p
Do you have any suggestions which blocks are responsible for hair?
missing these nodes, anybody know? can't find them:
workflowCHECK
workflow% Warmup Steps
Afdter the latest comfyUI-update it is slowed down extremely. If I used 3 hrs before it now will be more than 15 hrs.
For some reasons (that I have yet to understand) I couldn't use any Kohya_ss flux training on my pc, this workkflow allowed me to test lora training and worked great, but I would like to know, if you think you can do about the same kind of lora trainer but for hunyuan video (I mean on comfyui also) ?
I get the following error. Tried to mirror the settings you have but nodes have updated since those pictures.
'FluxNetworkTrainer' object has no attribute 'num_train_epochs'
After a comfy or node update from the author, selecting individual blocks for training stopped working. Do you have any suggestions on how this can be fixed?
Downloaded v1.5 today. Training 20 Images with default settings will take 9h on 3080 12G. Something wrong?
is there a step by step video with comentary to use this workflow thanks
Hi there..will this workflow work with the new Confyui update? Thanks
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.