| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Original HTTPS Page] |
Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.
You must be logged in to block users.
Contact GitHub support about this user’s behavior. Learn more about reporting abuse.
Report abuseorakul = {
"location" : "Chernihiv, Ukraine 🇺🇦",
"studio" : "Orakul Studio",
"stack" : ["Flux2", "ComfyUI", "PyTorch", "CUDA", "Python"],
"hardware" : "RTX 4090 · i9-13900K · 128GB RAM",
"specialty" : "AI Image Generation — No Quantization, No Compromise",
"stock" : "Adobe Stock · Macro · Scientific Photography",
"status" : "⚡ Working under fire. System stable.",
}
For most people, coding is a profession. A career. A way to make a living.
For me, code is a language.
Not a metaphor. It's literally the language I use to communicate with reality when words aren't enough. When you need to not describe a problem, but solve it. When you need to not complain about limitations, but break them.
I didn't go to MIT. I don't have a PhD in computer science. I have an RTX 4090, an i9-13900K, and a stubbornness that has outlasted everything the world has tried to do with it.
That was enough.High-performance LoRA training for Flux2 on RTX 4090.
Server-class speed on consumer hardware, The original ai-toolkit is an excellent, flexible framework. This fork takes it in one specific direction: maximum performance on RTX 4090 (Ada Lovelace, sm_89). Not about making weak hardware work. About making strong hardware fly.
BoneMemory: Universal Async Core for AI-Toolkit.
Whether you’re training a Rank 16 LoRA on an entry-level GPU or pushing Rank 1024 on an RTX 4090, BoneMemory eliminates memory bottlenecks.
Universal asynchronous VRAM engine for Ostris AI-Toolkit.
Zero-latency compute-transfer overlap with up to 11x training speedup on consumer GPUs.
Патч ядра PyTorch для RTX 4090 (sm_89).
Нативная генерация Flux2 2752×1536 без квантования за 5 минут.
Dual Export Node — PNG + 16-bit RAW TIFF из одной ноды ComfyUI.
Batch support · Accumulation mode · Zero compression loss
Макро-фотография насекомых и металлических текстур.
Adobe Stock · HDR · Black Background · Scientific Photography
🦟 Macro Entomology · 🔗 Rusty Metal · 🌀 AI Scientific Photography
Adobe Stock — 3rd place · Rusty Metal Chain series
«Уроборос кусает свой хвіст. І в цьому укусі — не смерть, а впізнавання.»
Orakul Protocol — кожен цикл мудріший за попередній.
Every iteration smarter than the last.
Fixed memory management logic for AI-Toolkit: Eliminates startup hangs and crashes through async CUDA stream optimization.
Python 1
Orakul SVP: The 16-bit Master Export Standard. Unlocking the full 16-bit potential of Flux and SDXL. Automatically accumulates your batch runs into a professional RAW TIFF library. Designed for tho…
Python 1
Unlocking RTX 4090: 15s/it for 56B models (Flux 2.0 Dev + Mistral 24B). Protocol Marianna F2.11, P-State 'Defibrillation' method & Workflow. (Перевод: Разблокировка RTX 4090: 15 сек/ит для моделей …
🚀 Advanced UniPC sampler surgical control node for ComfyUI. Exposing internal Predictor-Corrector mathematics to manage fine detail (D-tail) and noise trajectory. Precision sampling control for hig…
Python 1
A direct memory-to-RAM mapping protocol for PyTorch/CUDA. Bypasses the swap file/SSD page file for zero-latency AI computations. Optimized for RTX 4090/5090 and up to 128GB of RAM and higher.
Python 1
| Back | FazBrowse Home | New Git URL |