wan 2.2 lora training?
wan 2.2 lora training?
Nwill it include 14b parameter model @Furkan Gözükara SECourses
T@Furkan Gözükara SECourses are flux trainign in main branch of kohya_ss now?
Ayou mean qwen?
TMy scripts broke because sd3-flux.1 branch is removed now 
Twe used to use it
Fi recommend massed compute
Fit is installed there
Fjust run update and use
Fyes
Fyes for a long time
Fqwen image is in musubi tuner
Fi am making gui based on that
F
Fadding more info the gui settings right now
TI'ts been a while... do we need to manually reinstall torch and xformers still or is it fixed too in main?
CCant wait to see your reasult and advice.. Want to to my FINAL qwen and flux krea LORA or full model
CDr.Furkan, if you making GUI please add all new function like :
--timestep_sampling option with qwen_shift
--qinglong_qwen
---dynamo_backend INDUCTOR
etc..
I personally liked to use command but will be great and easier when we have GUI
--timestep_sampling option with qwen_shift
--qinglong_qwen
---dynamo_backend INDUCTOR
etc..
I personally liked to use command but will be great and easier when we have GUI
CWIll be great also see later optimal config files for example 16gb vram, 24gb vram etc...
CSO you thing my example pic are bad for some reason ?? Can you share you honest opinion ?
A@Furkan Gözükara SECourses when will you drop gui?
Dwhats the problem on runpod.. is OT not updated.. been using runpod and deposited what i had there.. i guess im not getting a refund but what update is missing
Fi think all in main
Fmy installer properly installs
Fcheck everyuthing and let me know what missing
Fstill improving gui adding more info
F
Acant see anything
Fi think training can become more realistic
Fdownload image
Fstill working
FOT is designed to have OS interface
Fthey added runpod like support later
Dohh so thats why.. damn...wish i knew this before paying for runpod
Fyep that is unfortunate they dont have browser gui
Ftheir gui also sucks so bad
CSince I've been deep in training with the command line, I had a few thoughts on what might be needed to get it fully working for models like Qwen .
The most important thing would be a dropdown at the top to choose the trainer script, like for qwen_image_train_network.py or flux_kontext_train_network.py, since they use diferent scripts.
In the Advanced tab, I definately needed options for the attention mechanism (--sdpa / --xformers) and the fp8_base / fp8_scaled toggles.
Those were super important to get my training running on a 24GB card without errors.
Also, the Network Module dropdown needs to have the specific options for each model, like networks.lora_qwen_image. My training failed until I found the right one for Qwen.
One small thing, maybe the precision setting (bf16/fp16) could be a dropdown insted of checkboxes?
Might be less confusing.
Just some ideas from my recent training hell lol.
The layout is great though, really clean. Keep up the awesome work!
The most important thing would be a dropdown at the top to choose the trainer script, like for qwen_image_train_network.py or flux_kontext_train_network.py, since they use diferent scripts.
In the Advanced tab, I definately needed options for the attention mechanism (--sdpa / --xformers) and the fp8_base / fp8_scaled toggles.
Those were super important to get my training running on a 24GB card without errors.
Also, the Network Module dropdown needs to have the specific options for each model, like networks.lora_qwen_image. My training failed until I found the right one for Qwen.
One small thing, maybe the precision setting (bf16/fp16) could be a dropdown insted of checkboxes?
Might be less confusing.
Just some ideas from my recent training hell lol.
The layout is great though, really clean. Keep up the awesome work!
Fi am simplifying
Fqwen tab only qwen 
COk. But always open gate to add more functionality later xd
F
F
F
CHere are the most critical things:
FP8 Options: We definately need two checkboxes in the "Advanced" tab: one for --fp8_base and one for --fp8_scaled. These were the biggest VRAM savers and the training wouldn't even start without them.
Blocks to Swap: Also in "Advanced", we need an input box for --blocks_to_swap. For a 1024x1024 training on a 24GB card, I had to use a value of 20 to prevent crashes.
Attention Mode: A dropdown to select the attention mechanism is also essential.
My training only worked after I set it to --sdpa.
Network Module: This one is critical! In the "LoRA Parameters" tab, the dropdown for Network Module absolutely needs to have networks.lora_qwen_image as an option. The generic networks.lora module made my training crash with an error.
One other small idea, maybe the Precision setting in the "Source Model" tab could be a dropdown menu (bf16, fp16) insted of separate checkboxes?
Might prevent users from accidentally clicking more than one.
For bonus points, adding fields for advanced stuff like LoRA+ (the network_args field you have is perfect for this!) and PyTorch Dynamo (--dynamo_backend) would make this the ultimate tool.
Hope this feedback from my command-line struggles is helpful! The GUI is already looking amazing. Can't wait to try it when it's ready. Keep up the awesome work!
FP8 Options: We definately need two checkboxes in the "Advanced" tab: one for --fp8_base and one for --fp8_scaled. These were the biggest VRAM savers and the training wouldn't even start without them.
Blocks to Swap: Also in "Advanced", we need an input box for --blocks_to_swap. For a 1024x1024 training on a 24GB card, I had to use a value of 20 to prevent crashes.
Attention Mode: A dropdown to select the attention mechanism is also essential.
My training only worked after I set it to --sdpa.
Network Module: This one is critical! In the "LoRA Parameters" tab, the dropdown for Network Module absolutely needs to have networks.lora_qwen_image as an option. The generic networks.lora module made my training crash with an error.
One other small idea, maybe the Precision setting in the "Source Model" tab could be a dropdown menu (bf16, fp16) insted of separate checkboxes?
Might prevent users from accidentally clicking more than one.
For bonus points, adding fields for advanced stuff like LoRA+ (the network_args field you have is perfect for this!) and PyTorch Dynamo (--dynamo_backend) would make this the ultimate tool.
Hope this feedback from my command-line struggles is helpful! The GUI is already looking amazing. Can't wait to try it when it's ready. Keep up the awesome work!
F
CYou think about everything
CBoss
CI learned a lot from you btw.
Jnetwork dim , network alpha, vae_cache cpu?
Jnext thing you know , you rewrote koyha_ss gui 
Athis is just amazing
