bmaltais / kohya_ss

Repository from Github https://github.combmaltais/kohya_ss

bmaltais/kohya_ss Issues

Can't load SD 3.5 checkpoint // sd3-sd3.5-flux branch
Updated a month ago1
Lora Training ending with monitor refresh?
Closed a month ago2
Stuck to : " INFO Loading settings from "
Updated 2 months ago5
use Blocks to swap error
Updated 2 months ago2
Kohya works fine with SD1.5 but crashes with SDXL ?!
Updated 2 months ago1
Someone could help me? returned non-zero exit status 1.
Updated 2 months ago3
Flux Dreambooth/finetune 是否能在4090 24GVRAM上成功训练？
Updated 2 months ago5
I was doing flux finetune, training based on flux1.dev base model, and then encountered time_embed.0.weight when extract lora.
Closed 2 months ago1
"the following directories listed in your path were found to be non-existent" but they are not in any path.
Updated 2 months ago
LD_LIBRARY_PATH environment variable is not set (linux question)
Updated 2 months ago2
Returned non-zero exit status 3221225477
Updated 2 months ago2
ERROR No data found. Please verify arguments (train_data_dir must be the train_network.py:212 parent of folders with images) /
Updated 2 months ago1
LoRa training with no results
Updated 2 months ago1
TypeError: int() argument must be a string, a bytes-like object or a real number, not 'NoneType'
Updated 2 months ago1
fail to start training, anyone help?
Updated 2 months ago5
Can you add support for sd3.5 lora training, its available on sd scripts
Updated 2 months ago2
is accelerate called twice on a 3090 ?
Updated 2 months ago
LORA merge SDXL sometimes not work
Updated 2 months ago
SDXL training stops after generate first sample image
Updated 2 months ago
problem with textual inversion tab
Updated 2 months ago
[FEATURE] - Florence2 Captioning
Updated 2 months ago
Flux LoRA and OOM on a 4090?
Updated 2 months ago4
Kohya and Flux resulting in access violation error (returned non-zero exit status 3221225477)
Closed 2 months ago1
jxl support
Updated 2 months ago
BLIP2 captioning broken
Updated 2 months ago
Training LoRA from fine-tuned model fails at loading t5xxl_fp16.safetensors
Updated 2 months ago2
ValueError: bypass mode is not supported in Full algo.
Updated 2 months ago
Resume from saved training state not working - error
Closed 2 months ago1
Manual Captioning (editing) with latest version, with the updated gradio now doesn't work properly/as it did before
Updated 2 months ago
In the section for loading and saving a Lora or Dreambooth config file there's no "Save as" but there used to be
Updated 2 months ago
flux lora merger only works with cpu device
Updated 2 months ago
Can we use Differential Output Preservation loss for LoRA at the moment
Updated 2 months ago9
Not every Flux Lora can be merged with Kohya
Updated 2 months ago2
Step calculation is broken - 200 epoch - 28 training images and it makes 1600 steps - FLUX branch
Updated 2 months ago10
Google Colab Notebook: ImportError: cannot import name 'split_torch_state_dict_into_shards' from 'huggingface_hub' (/usr/local/lib/python3.10/dist-packages/huggingface_hub/__init__.py)
Updated 2 months ago1
Incorrect calculation of “Max train steps”.
Updated 2 months ago8
the state of max grad norm is not saved in the json configuration file
Updated 2 months ago
ERROR: UnboundLocalError: local variable 'max_train_steps_info' referenced before assignment
Updated 2 months ago8
Configuration File switch dreambooth/lora tab
Updated 2 months ago
Google Colab Notebook error: ModuleNotFoundError: No module named 'torch._custom_ops'
Closed 2 months ago
update pytorch windows dependencies in dev branch
Updated 2 months ago1
Blocks to swap for flux and sd3.5 lora? its available
Updated 2 months ago17
SafetensorError: Error while deserializing header: InvalidHeaderDeserialization when open .safetensor model
Updated 2 months ago
Which branch should I switch to when training flux lora? There are three branches of flux and I don't know which one to switch to?
Updated 3 months ago2
Should I be worried? My training counters are very high.
Updated 3 months ago
Fix dependencies for linux in sd3-flux.1 and sd3-sd3.5-flux branches
Closed 3 months ago13
no working Pony/SDXL
Updated 3 months ago
MileStones LR scheduler extra arguments can not use
Closed 3 months ago3
torch.OutOfMemoryError: CUDA out of memory.
Updated 3 months ago
Request: Add Support for V-Prediction SDXL Model in LoRA Training
Closed 3 months ago1