bmaltais/kohya_ss issues and pull requests

#2892 - Can we perform flux finetunes and loras using the q8 (gguf) quantized version?

Issue - State: open - Opened by Vigilence about 1 month ago

#2891 - Deprecation. cant install the kohya.

Issue - State: open - Opened by TheWorl4 about 1 month ago - 1 comment

#2890 - Finetunning Issue with meta_lat.json in FLUX Kohya.

Issue - State: open - Opened by davizca about 1 month ago - 2 comments

#2889 - Dreambooth Error [FLUX]

Issue - State: closed - Opened by Crimsonfart about 1 month ago

#2888 - Sizes of tensors must match except in dimension 1

Issue - State: open - Opened by jigrain about 1 month ago - 1 comment

#2887 - [Bug] `subprocess.CalledProcessError` occurred while running `train_network.py` script

Issue - State: open - Opened by baicai99 about 1 month ago - 1 comment

#2886 - Why it must be:TensorBoard at http://0.0.0.0:6006/

Issue - State: open - Opened by hben35096 about 1 month ago

#2885 - nyanko7/flux-dev-de-distill model support

Issue - State: open - Opened by dsienra about 1 month ago - 1 comment

#2884 - Console feedback on use of masked loss, and a question

Issue - State: closed - Opened by os144046 about 1 month ago - 2 comments

#2883 - --help is ignored when --inbrowser used - I want to prevent requirements check

Issue - State: closed - Opened by FurkanGozukara about 1 month ago - 6 comments

#2882 - training for lora model keeps ending

Issue - State: open - Opened by lier90912 about 1 month ago - 1 comment

#2880 - BUG: sample images - VRAM overflow

Issue - State: open - Opened by kalle07 about 1 month ago

#2876 - Fix typo on README.md

Pull Request - State: closed - Opened by bulieme about 1 month ago

#2875 - Bump crate-ci/typos from 1.23.6 to 1.25.0

Pull Request - State: closed - Opened by dependabot[bot] about 1 month ago
Labels: dependencies, github_actions

#2874 - After upgrading and rerunning the setup, Flux finetuning from 14s/it into 164s/it

Issue - State: open - Opened by veighnsche about 1 month ago - 1 comment

#2872 - AdEMAMix8bit settings for FLUX

Issue - State: open - Opened by pAInCREAT0R about 1 month ago

#2871 - Flux Dreambooth training broke sometime after commit sd3-flux.1 95bf7ff29cade27108a55c49f36f360c7449cde0

Issue - State: closed - Opened by mostlyhuman about 1 month ago - 6 comments

#2870 - Is Multi GPU working yet with Flux.1 Dreambooth?

Issue - State: open - Opened by mostlyhuman about 2 months ago - 1 comment

#2869 - TypeError: ForwardRef.init() got an unexpected keyword argument 'is_class'

Issue - State: open - Opened by DarkPulse about 2 months ago

#2868 - feat(docker): mount models folder as a volume

Pull Request - State: closed - Opened by vvsotnikov about 2 months ago

#2867 - ⚡Align GUI with Flux-Dev2Pro Support from sd-scripts

Issue - State: open - Opened by sashaok123 about 2 months ago - 1 comment

#2866 - Min Timestep and Max Timestep

Issue - State: open - Opened by jigrain about 2 months ago

#2860 - Solving Class Bleeding Post-Distillation in Flux1-Dev with Full Fine-Tuning and LoRA

Issue - State: open - Opened by dsienra about 2 months ago

#2859 - Can we have model conversion script for FLUX? FP16 to FP8

Issue - State: open - Opened by FurkanGozukara about 2 months ago - 4 comments

#2858 - TAESD-preview on CPU?

Issue - State: open - Opened by iqddd about 2 months ago

#2857 - Linux Docker Training Error: Validating /app/data/img existence... FAILED: does not exist lora training

Issue - State: open - Opened by shalini8426222 about 2 months ago

#2856 - after 1epoch,the trainning is interrunpt,can not load the pre-trained data

Issue - State: open - Opened by Lecho303 about 2 months ago - 2 comments

#2855 - LoRA Training fails with ponyDiffusionV6XL Model

Issue - State: closed - Opened by samuelkurt about 2 months ago - 2 comments

#2854 - images in TensorBoard

Issue - State: open - Opened by titan0115 about 2 months ago

#2853 - Network parameters missing

Issue - State: closed - Opened by lynrayy about 2 months ago - 4 comments

#2852 - subprocess.CalledProcessError: Command '['C:\\Users\\18078\\AppData\\Local\\Programs\\Python\\Python310\\python.exe', 'D:/kohya/kohya_ss/sd-scripts/train_network.py', '--config_file', 'D:\\kohya\\kohya_ss\\dataset\\huangquan\\output/config_lora-20240924-115523.toml']' returned non-zero exit status 1.

Issue - State: open - Opened by AlexiosDyral about 2 months ago - 6 comments

#2851 - Allow LoKr training on Flux

Issue - State: open - Opened by dalmackay about 2 months ago

#2850 - Got 2 errors when training SDXL including a non-zero exit status 1

Issue - State: open - Opened by skelescene about 2 months ago - 1 comment

#2849 - Overwrite pop up error on RunPod - no desktop UI - sd3-flux.1

Issue - State: open - Opened by FurkanGozukara about 2 months ago - 1 comment

#2848 - Update requirements_pytorch_windows.txt

Pull Request - State: closed - Opened by wzgrx about 2 months ago

#2847 - Why should version mandatory verification be enabled by default in 2024? Strongly recommend canceling the verification.

Issue - State: open - Opened by wzgrx about 2 months ago

#2846 - libuv errors when attempting multi-gpu dreambooth training in sd3-flux.1 branch

Issue - State: open - Opened by mostlyhuman about 2 months ago - 1 comment

#2845 - Loaded the config file, but the parameters and args remain the same

Issue - State: closed - Opened by RaySteve312 about 2 months ago - 1 comment

#2844 - Multi GPU training checkpoint problem

Issue - State: open - Opened by lixida123 about 2 months ago - 1 comment

#2843 - returned non-zero exit status 3221225477.

Issue - State: open - Opened by max-k54 about 2 months ago - 3 comments

#2842 - BUG: FLUX Lora merge, no stop option, no save option

Issue - State: open - Opened by kalle07 about 2 months ago

#2841 - TypeError: get_cosine_schedule_with_warmup() got an unexpected keyword argument 'num_decay_steps'

Issue - State: open - Opened by AndreR about 2 months ago - 1 comment

#2840 - Feature: merge flux, merge anyway

Issue - State: open - Opened by kalle07 about 2 months ago

#2839 - Flux merge tool does not properly handle LoRA keys from AI-Toolkit when merging w/ Flux.dev model - (Kohya responded on their repo saying flag needed - can toggle be added?)

Issue - State: open - Opened by CCpt5 about 2 months ago

#2838 - How to save last lora

Issue - State: open - Opened by Chengym2023 about 2 months ago

#2837 - Epochs is locked to 8 when using multires dataset config

Issue - State: open - Opened by dalmackay about 2 months ago

#2836 - Extract Flux LoRA feature is not working

Issue - State: closed - Opened by RayHell about 2 months ago - 1 comment

#2834 - LR warmup error AttributeError: 'BasicTraining' object has no attribute 'old_lr_warmup_steps'. Did you mean: 'lr_warmup_steps'?

Issue - State: open - Opened by Kikunu about 2 months ago - 1 comment

#2833 - Latest editions of the gui issues.

Issue - State: closed - Opened by DarkAlchy about 2 months ago

#2830 - Fixes for Mac M3

Pull Request - State: open - Opened by JoeyOverby about 2 months ago - 8 comments

#2829 - Update 3 files in kohya_gui folder

Pull Request - State: closed - Opened by iamrohitanshu about 2 months ago

#2828 - "index_select_cuda" not implemented for 'Float8_e4m3fn'

Issue - State: closed - Opened by DarkAlchy about 2 months ago

#2827 - Can't use GIT PULL

Issue - State: open - Opened by Arnold2006 about 2 months ago - 2 comments

#2826 - Flux concept bleeding.

Issue - State: open - Opened by dsienra about 2 months ago - 2 comments

#2825 - flux lora training does not start,error: returned non-zero exit status 1

Issue - State: open - Opened by zxf5843029 about 2 months ago

#2824 - Training has ended

Issue - State: open - Opened by jigrain about 2 months ago - 1 comment

#2823 - TypeError: ClusterConfig.init() got an unexpected keyword argument 'tpu_use_cluster'

Issue - State: open - Opened by noya21th 2 months ago

#2822 - console Log INFO always shows filepath

Issue - State: closed - Opened by harahara777 2 months ago - 4 comments

#2821 - Feature: fp8 data save

Issue - State: open - Opened by kalle07 2 months ago

#2820 - "Cannot copy out of meta tensor; no data!" right at the beginning of the training

Issue - State: open - Opened by AndreR 2 months ago - 9 comments

#2819 - feature: flux training 16GB with 1024pix

Issue - State: closed - Opened by kalle07 2 months ago

#2817 - cant start training File "D:\KohyaNew\kohya_ss\kohya_gui\lora_gui.py", line 994, in train_model if max_train_steps > 0 and stop_text_encoder_training > 0: TypeError: '>' not supported between instances of 'NoneType' and 'int'

Issue - State: open - Opened by DarkViewAI 2 months ago - 5 comments

#2814 - "No module named 'global_state'" in gui.bat

Issue - State: open - Opened by AndreR 2 months ago - 1 comment

#2813 - FLUX Merge

Issue - State: closed - Opened by kalle07 2 months ago

#2812 - got an unexpected keyword argument 'num_decay_steps'

Issue - State: open - Opened by Nova-lotus 2 months ago - 9 comments

#2811 - HELP - bitsandbytes.optim.adamw.AdamW8bit(weight_decay=0.01,eps=1e-08,betas=(0.9, 0.999))

Issue - State: closed - Opened by MIKETUFFIN66 2 months ago - 5 comments

#2810 - WSL, Torch reports GPU not available on AMD

Issue - State: open - Opened by Charmandrigo 2 months ago

#2809 - Multi Error Install Kohya_ss Flux

Issue - State: open - Opened by XT-404 2 months ago

#2808 - feature: other learning option (LR Scheduler)

Issue - State: open - Opened by kalle07 2 months ago

#2807 - Error during Git operation: Command '['git', 'submodule', 'update', '--init', '--recursive', '--quiet']' returned non-zero exit status 128.

Issue - State: closed - Opened by MIKETUFFIN66 2 months ago - 4 comments

#2806 - Runway removed repo, so 1.5 training will not download model

Issue - State: open - Opened by revolvedai 2 months ago - 2 comments

#2805 - Error: NSWindow should only be instantiated on the main thread!

Issue - State: open - Opened by JoeyOverby 2 months ago - 1 comment

#2804 - resize lora flux - slow

Issue - State: open - Opened by kalle07 2 months ago - 2 comments

#2803 - docker fails

Issue - State: open - Opened by fahadshery 2 months ago - 2 comments

#2802 - Add dark mode css variables

Pull Request - State: closed - Opened by wcole3 2 months ago - 1 comment

#2801 - Sample prompt can't be trusted - adds in extra tokens

Issue - State: open - Opened by JoeyOverby 2 months ago

#2800 - Returned non-zero exit status 3221225477 - can't train lora

Issue - State: open - Opened by Telllinex 2 months ago - 12 comments

#2799 - New Installation Promlem

Issue - State: closed - Opened by chnisar515 2 months ago - 2 comments

#2798 - AttributeError: 'LoRANetwork' object has no attribute 'train_blocks'

Issue - State: open - Opened by SailingB0aty 2 months ago - 1 comment

#2797 - **Important (Timestep Issue)

Issue - State: open - Opened by DriveHabits 2 months ago - 17 comments

#2796 - I believe flux_shift timestep is not working correctly

Issue - State: closed - Opened by DriveHabits 2 months ago

#2795 - Cannot train: kohya_ss on branch sd3-flux.1 on latest version causes training to stop and freeze

Issue - State: open - Opened by bolli20000 2 months ago - 3 comments

#2794 - Can not find TensorRT and CudNN when they are installed for kohya_ss-sd3-flux.

Issue - State: open - Opened by SkyHoshino 2 months ago - 5 comments

#2793 - Multi resolution training for Flux

Issue - State: open - Opened by nitinh12 2 months ago - 6 comments

#2791 - ERROR: Could not install packages due to an OSError: [WinError 5] Check the permissions.

Issue - State: open - Opened by plugcrypt 2 months ago

#2790 - Connection errored out after install kohya.

Issue - State: closed - Opened by Salami50 2 months ago - 5 comments

#2789 - BUG: FluX error train stops since today update

Issue - State: closed - Opened by kalle07 2 months ago - 12 comments

#2788 - I get a connection error whenever I click on any button

Issue - State: closed - Opened by Philip0909 2 months ago

#2787 - Loaded runtime CuDNN library

Issue - State: open - Opened by phucbienvan 2 months ago

#2785 - Need urgent fastapi fix

Issue - State: open - Opened by FurkanGozukara 2 months ago - 4 comments

#2784 - After uninstalling old version to try to get the flux training going I get connection timed out error

Issue - State: closed - Opened by PheonixAi420 2 months ago - 2 comments

#2783 - PydanticSchemaGenerationError for class starlette.requests.Request

Issue - State: open - Opened by Mark-Reiser 2 months ago - 12 comments

#2782 - BUG: FLUX branch STOP Button

Issue - State: open - Opened by kalle07 2 months ago - 1 comment

#2781 - [BUG] Flux Lora training - sample image

Issue - State: closed - Opened by kalle07 2 months ago - 2 comments

#2780 - [BUG] Typo for GUI timestep_sampling flux_shift

Issue - State: open - Opened by hinablue 2 months ago - 4 comments

#2779 - ValueError: T5XXL is trained, so cache_text_encoder_outputs cannot be used /

Issue - State: open - Opened by DarkViewAI 2 months ago

#2778 - question: is lora and checkpoint merge possible for FLUX ?

Issue - State: open - Opened by kalle07 2 months ago - 2 comments

#2777 - Can you check these possible errors with t5 attention mask and flux shift timestep and Discrete Flow Shift

Issue - State: open - Opened by DarkViewAI 2 months ago

#2776 - feature: downloaded models

Issue - State: closed - Opened by kalle07 2 months ago

#2775 - Any chance we can force accelerate to use cuda?

Issue - State: closed - Opened by FurkanGozukara 2 months ago

GitHub / bmaltais/kohya_ss issues and pull requests