Qwen Image Edit Plus 2509 Model full Fine Tuning and LoRA training speeds and VRAM usages on RTX 6000 PRO GPU with absolute maximum quality and 1328x1328 pixel training resolution - 1 Control image makes it like 100% slower compared to no Control images

Made extremely important 4 pull requests to Kohya Musubi Tuner repo I hope all gets implemented / merged asap

Qwen Image Edit Plus 2509 model full fine tuning RTX 5090 speed on RunPod with 1328x1328px resolution + control images - base model without control images almost 2x speed

Qwen Image 1328x1328px 128 Rank LoRA training speed on Windows RTX 5090 - Currently sadly Windows is slow due to most of the time being spent on RAM to GPU and GPU to RAM - This doesnt happen on Linux I have notified Kohya lets see if he can improve Block Swapping

Uploading Qwen Image training R&D test model checkpoints to Hugging Face

RTX 5090 Qwen Image LoRA training Speed are like 3.85 second / it for FP8 Scaled and 4.5 second / it for BF16 - 128 Rank - 1328px - For Linux only sadly - Windows still much slower due to Block Swap GPU - RAM synch issue

SECourses Premium Kohya GUI V32 Update has arrived - we have up-to-date FLUX Training DreamBooth / Fine Tuning and LoRA training presets as well - Like including FLUX SRPO

People loving voice translation feature of YouTube I hope YouTube Liaison makes it more accurate. Because YouTube transcription is like 100x worse than transcribing with Gemini

Qwen Image Models (Base + Edit) training (LoRA + Fine Tuning) tutorial will be published hopefully tomorrow on SECourses YouTube channel. This tutorial is around 2 hours and itm is a total masterpiece that is like a full course.

Qwen Image models full Fine Tuning speed on Windows with RTX 5090

I have made topics of our all videos on our GitHub including with transcriptions / subtitles Now I can search the words spoken in videos to find them Helping amazingly with index. Because on YouTube there is no such feature

Meanwhile our newest tutorial published : youtu.be/MLNQkwW_Q8M New Qwen Image Edit Fusion, Re-Light and ReAngle / Multiple Angles LoRAs Tutorial - Easy and Fast

Updating our Model downloadar app for FLUX 2 models. Every App using different folder names :D Supporting SwarmUI, ComfyUI, Forge Web UI Classic and Forge Web UI Neo. Hopefully will make a tutorial + presets for FLUX 2 very soon

My GitHub profile as of 20 December 2025 (Furkan Gözükara)