Home Creators Posts Import Register Favorites Logout

Content

FLUX 2 Pro compared with a trained FLUX SRPO model, plus the new Kohya trainer V35 that trains FLUX on GPUs with as little as 6 GB VRAM.

The Kohya SS GUI Premium app gets updated presets and new features: full Torch Compile, on-the-fly FP8 scaled LoRA training, an 80 GB GPU config, a FLUX FP8 converter, a dataset pre-processing preview and an automatic model downloader. Covers LoRA and DreamBooth fine tuning.

Tutorial Video Link

https://youtu.be/RQHmyJVOHXo

Info

FLUX 2 has been published and I have compared it to the very best FLUX base model known as FLUX SRPO. Moreover, we have updated our FLUX Training APP and presets to the next level. Massive speed up gaings with 0 quality loss and lots of new features. I will show all of the new features we have with new SECourses Kohya SS GUI Premium app and compare FLUX SRPO trained model results with FLUX 2.

Resources

Some FLUX SRPO Generations With Medium Quality Training Images Dataset (28) - Below Same Prompts FLUX 2 PRO

This below image blocked on BFL site so only exists in local image as comparison

Some FLUX 2 PRO Generations - BFL Website Playground Used - 2048x1152 px

Official BFL Playground Link

  • https://playground.bfl.ai/image/generate

⏱️ Video Chapters:

  • 0:00 Introduction to New FLUX Training Improvements and Local Training Showcase

  • 0:24 Understanding FLUX SRPO Model: High Realism with Minimal VRAM Requirements

  • 0:38 Updated Configurations for Training Realism on 6GB VRAM GPUs Locally

  • 1:07 FLUX 2 Announcement and Setting Up Comparisons with BFL Playground

  • 1:45 FLUX 2 Dev Model Technical Specs: 32 Billion Parameters and Hardware Challenges

  • 2:11 Overview of Changes in SECourses Premium Kohya Trainer Version 35

  • 2:46 Development Updates: GUI Improvements and Full Torch Compile Support

  • 3:13 LoRA Presets Update: VRAM Optimization and Speed Improvements via Torch Compile

  • 3:27 Introducing On-the-Fly FP8 Scaled LoRA Training Support

  • 3:42 Quality Comparison Analysis: BF16 vs FP8 Scaled Weights LoRA

  • 4:24 VRAM Usage and Speed Analysis: Block Swap Count Reduction with FP8 Scaled

  • 5:12 Why 32GB GPUs Don't Need FP8 Scaled: Fitting Completely with Torch Compile

  • 5:36 DreamBooth Training Specifics: BF16 Mixed Precision and Tab Selection

  • 6:18 New 80GB GPU Configuration: Significant Training Speed Up and Cost Analysis

  • 7:39 New Feature: FLUX FP8 Converter Tool for DreamBooth Models

  • 8:10 Verifying Quality of Converted FP8 Scaled Models vs Original BF16

  • 9:03 New Image Pre-processing Tool: Visualizing How Kohya Sees Your Dataset

  • 9:34 Demonstration of Pre-processing: Identifying Padding and Orientation Issues

  • 10:38 Additional Features: Memory Efficient Loading and CPU Text Encoder Caching

  • 11:23 New Automated Model Downloader Tool: Installation and Model Selection Guide

  • 12:35 FLUX 2 vs FLUX 1 Context: Can the New Model Replace Existing Workflows?

  • 13:03 Setting Up the Generation Comparison: SRPO Fine-tune vs Base vs FLUX 2 Pro

  • 14:13 Analyzing First Comparison Results: Portrait Quality and Realism Assessment

  • 15:02 Second Comparison Test: Handling Unrealistic Elements and Animals in Prompts

  • 15:33 Importance of Resolution and Platform Choice for FLUX 2 Quality

  • 16:24 Analyzing Second Results: Prompt Adherence and Realism in FLUX 2 vs SRPO

  • 17:07 Testing the Prompt on Nano Banana Pro: Quality Assessment

  • 17:39 Comparison with Seedream 4 Model and Final Thoughts on FLUX 2 Potential

  • 18:50 Current Recommendation: Why Qwen Image Realism is the Temporary King

More Info

  • In this video, I demonstrate the newest improvements and features added to our SECourses Premium Kohya Trainer (v35). We have achieved massive performance gains and VRAM reductions, allowing for high-quality FLUX training on GPUs with as little as 6GB of VRAM using the FLUX SRPO model.

  • I also break down the brand new FLUX 2 announcement! We perform a side-by-side comparison between my locally trained FLUX SRPO model, the base FLUX 1 Dev, and the newly released FLUX 2 Pro model to see if it’s time to switch workflows.

🚀 Key Updates Covered:

  • Full Torch Compile Support: Now enabled across all presets for faster training speeds.

  • FP8 Scaled LoRA Training: On-the-fly conversion that drastically reduces VRAM usage (0 block swaps on 24GB cards) with zero quality loss.

  • New Tools: Introducing the "FLUX FP8 Converter" to shrink DreamBooth models to 11.1GB and a new "Image Pre-processing" tool to visualize exactly how Kohya sees your dataset.

  • Hardware Optimization: New presets ranging from 6GB VRAM consumer cards up to 80GB A100/H100 configurations for lightning-fast training.

  • FLUX 2 Analysis: A realistic look at the 32-billion parameter giant and how it compares to our optimized FLUX 1 workflow.

Related posts

All posts in this topic: LoRA and Fine-Tuning Training: Kohya, Musubi Tuner, OneTrainer, AI Toolkit · New to SECourses? Start Here: Best and Most Popular SECourses Posts

Files

FLUX 2 vs FLUX SRPO, New FLUX Training Kohya SS GUI Premium App With Presets & Features

FLUX 2 has been published and I have compared it to the very best FLUX base model known as FLUX SRPO. Moreover, we have updated our FLUX Training APP and presets to the next level. Massive speed up gaings with 0 quality loss and lots of new features. I will show all of the new features we have with new SECourses Kohya SS GUI Premium app and compare FLUX SRPO trained model results with FLUX 2. Get the SECourses Premium Kohya Trainer DreamBooth / Fine Tuning : [ https://www.patreon.com/posts/Kohya-FLUX-DreamBooth-Trainer-App-112099700 ] Get the SECourses Premium Kohya Trainer LoRA : [ https://www.patreon.com/posts/Kohya-FLUX-LoRA-Trainer-App-110879657 ] DreamBooth Training Tutorial: [ https://www.youtube.com/watch?v=FvpWy1x5etM ] LoRA Training Tutorial: [ https://www.youtube.com/watch?v=nySGu12Y05k ] Qwen Image Realism Tutorial: [ https://youtu.be/XWzZ2wnzNuQ ] Join our Discord Community: [ https://discord.com/servers/secourses-Discord-772774097734074388 ] ⏱️ Video Chapters: 0:00 Introduction to New FLUX Training Improvements and Local Training Showcase 0:24 Understanding FLUX SRPO Model: High Realism with Minimal VRAM Requirements 0:38 Updated Configurations for Training Realism on 6GB VRAM GPUs Locally 1:07 FLUX 2 Announcement and Setting Up Comparisons with BFL Playground 1:45 FLUX 2 Dev Model Technical Specs: 32 Billion Parameters and Hardware Challenges 2:11 Overview of Changes in SECourses Premium Kohya Trainer Version 35 2:46 Development Updates: GUI Improvements and Full Torch Compile Support 3:13 LoRA Presets Update: VRAM Optimization and Speed Improvements via Torch Compile 3:27 Introducing On-the-Fly FP8 Scaled LoRA Training Support 3:42 Quality Comparison Analysis: BF16 vs FP8 Scaled Weights LoRA 4:24 VRAM Usage and Speed Analysis: Block Swap Count Reduction with FP8 Scaled 5:12 Why 32GB GPUs Don't Need FP8 Scaled: Fitting Completely with Torch Compile 5:36 DreamBooth Training Specifics: BF16 Mixed Precision and Tab Selection 6:18 New 80GB GPU Configuration: Significant Training Speed Up and Cost Analysis 7:39 New Feature: FLUX FP8 Converter Tool for DreamBooth Models 8:10 Verifying Quality of Converted FP8 Scaled Models vs Original BF16 9:03 New Image Pre-processing Tool: Visualizing How Kohya Sees Your Dataset 9:34 Demonstration of Pre-processing: Identifying Padding and Orientation Issues 10:38 Additional Features: Memory Efficient Loading and CPU Text Encoder Caching 11:23 New Automated Model Downloader Tool: Installation and Model Selection Guide 12:35 FLUX 2 vs FLUX 1 Context: Can the New Model Replace Existing Workflows? 13:03 Setting Up the Generation Comparison: SRPO Fine-tune vs Base vs FLUX 2 Pro 14:13 Analyzing First Comparison Results: Portrait Quality and Realism Assessment 15:02 Second Comparison Test: Handling Unrealistic Elements and Animals in Prompts 15:33 Importance of Resolution and Platform Choice for FLUX 2 Quality 16:24 Analyzing Second Results: Prompt Adherence and Realism in FLUX 2 vs SRPO 17:07 Testing the Prompt on Nano Banana Pro: Quality Assessment 17:39 Comparison with Seedream 4 Model and Final Thoughts on FLUX 2 Potential 18:50 Current Recommendation: Why Qwen Image Realism is the Temporary King In this video, I demonstrate the newest improvements and features added to our SECourses Premium Kohya Trainer (v35). We have achieved massive performance gains and VRAM reductions, allowing for high-quality FLUX training on GPUs with as little as 6GB of VRAM using the FLUX SRPO model. I also break down the brand new FLUX 2 announcement! We perform a side-by-side comparison between my locally trained FLUX SRPO model, the base FLUX 1 Dev, and the newly released FLUX 2 Pro model to see if it’s time to switch workflows. 🚀 Key Updates Covered: Full Torch Compile Support: Now enabled across all presets for faster training speeds. FP8 Scaled LoRA Training: On-the-fly conversion that drastically reduces VRAM usage (0 block swaps on 24GB cards) with zero quality loss. New Tools: Introducing the "FLUX FP8 Converter" to shrink DreamBooth models to 11.1GB and a new "Image Pre-processing" tool to visualize exactly how Kohya sees your dataset. Hardware Optimization: New presets ranging from 6GB VRAM consumer cards up to 80GB A100/H100 configurations for lightning-fast training. FLUX 2 Analysis: A realistic look at the 32-billion parameter giant and how it compares to our optimized FLUX 1 workflow.

Comments

EvanGuo

及时优秀的工作!请问您知道chroma的lora训练吗

SECourses

yes i plan chroma research later hopefully

EvanGuo

flux pro和flux2 的preset文件在哪

SECourses

yes i did huge search on this. will publish tomorrow hopefully sorry for delay

Roni Bliss

Vs Z-Image ... Flux 2 was just murdered in cold blood. All blck forests years of work down the drain. Flux 2 isnow the new SD 3.5 cast into the shadows a slow clunky relic. WOW

100Sachen80

Will Flux2 be trainable as Lora or Full Checkpoint with your Kohya Training App?

SECourses

yes hopefully with musubi tuner app

Hugo Morosini

i'm having low likeness doing a flux dev finetune with "80GB_GPUs_60GB_VRAM_Quality_1.json". I'm not changing any parameters beside the model paths, is there anything I should enable / disable ? I've noticed many things such as "fused_backward_pass" are now disabled, is that normal ? Thanks

Hugo Morosini

my dataset is good, and i've used it before using your previous tutorial and I obtained high likeness. However i'm using sd-scripts directly without using khoya GUI. It kind of work but it's like it's learning very very slowly and it looks 5% like my subject (I can see removing the lora change the face). I'm doing the 80GB training in a RTX Pro 6000 on runpod. I keep trying but it simply doesn't seem to learn. Loss is stuck at around ~0.4, is that normal ? Thanks

Hugo Morosini

Another question I had is have you had any success using Prodigy as an optimizer ? Thanks

Ec Jep

deleting this folder fixed the issue: AppData\Local\Temp\torchinductor

SECourses

thanks for info i will tell others too

SECourses

sadly i dont know. i recommend use gui and our configs

SECourses

i tested it extensively and sadly not

EvanGuo

H100训练,结果lora居然有2g大小,什么原因呢

Ec Jep

do your latest presets in koyhaV41 work with finetune training the flux2-dev-fp8mixedfromscaled model? (on 4090 24gb vram and 124 ram?

SECourses

I am adding flux 2 support currently not done yet

SECourses

English please but h100 works great

Ran Bensimon

will it work with Klein too?

SECourses

yes i just updated the app yesterday please read the changelogs now we have full klein support