Skip to content
Images

Fibo 2.0 Edit

IDM 2-Steps: LoRA для редактирования изображений в ComfyUI за 2 шага

2 min read

Это ComfyUI-конвертация LoRA akhaliq/Fibo-2.0-edit-IDM-2steps от NidAll. LoRA запускает редактирование изображений Bria Fibo 2.0 Edit Turbo за 2 шага без CFG и обучена на community-реализации Intrinsic Distribution Matching (IDM). В этом репозитории изменена только раскладка файлов, чтобы нативная модель Fibo 2 в ComfyUI могла загрузить LoRA.

Ключевые особенности

  • LoRA в формате ComfyUI: diffusion_model.*, lora_down/lora_up/alpha, bf16.
  • Применяется через LoraLoaderModelOnly со strength 1.0; сэмплирование — 2 шага, CFG 1.0.
  • Слои diffusers слиты под ComfyUI: qkv, to_kv, to_out, w13; rank 192 для qkv и 128 для to_kv/w13.
  • В репозитории есть скрипт конвертации convert_fibo2_lora.py.
  • Подходит только для ComfyUI-чекпоинтов с fused-слоями; для diffusers и чекпоинтов с раздельными w1/w3 или to_q/to_k/to_v нужен оригинальный LoRA.

Пример

Сценарий в ComfyUI: положите Fibo-2.0-edit-IDM-2steps_comfy.safetensors в ComfyUI/models/loras/, загрузите Fibo 2.0 Edit Turbo в fused-раскладке (например, fibo2_edit_turbo_int8_convrot.safetensors или fibo2_edit_turbo_bf16.safetensors из Comfy-Org/Fibo-2.0), примените LoRA через LoraLoaderModelOnly со strength 1.0 и сэмплируйте 2 шага с CFG 1.0.

ПодробнееФайл: 797 MB против 539 MB у оригинала — больше из-за zero-padding.
LoRA: bf16; alpha выставлена равной rank, поэтому масштаб остаётся 1.0, как в оригинале (r=64, alpha=64).
Лицензия: Bria Fibo terms, CC BY-NC 4.0 — только некоммерческое использование; для коммерции нужно отдельное соглашение с Bria AI.
Ограничение: не подходит для чекпоинтов, где w1/w3 или to_q/to_k/to_v остаются отдельными слоями.
Детали обучения, evals и сэмплы — в оригинальном репозитории.
Discuss Telegram 0

Will it run?

Working out the memory needs

This is a LoRA: it only runs on top of the base model akhaliq/Fibo-2.0-edit-IDM-2steps. The base model's size is still being collected; the estimate will appear within a day.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # NidAll/Fibo-2.0-edit-IDM-2steps-ComfyUI
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download akhaliq/Fibo-2.0-edit-IDM-2steps --local-dir ~/models/Fibo-2.0-edit-IDM-2steps
    hf download NidAll/Fibo-2.0-edit-IDM-2steps-ComfyUI --local-dir ~/models/Fibo-2.0-edit-IDM-2steps-ComfyUI
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (3)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • Nunchaku (SVDQuant 4-bit)v1.2.1from January 25, 2026

    ≈3× faster and 3.5× less memory

    4-bit versions of popular image models with almost no quality loss (FP4 on RTX 50).

    How to enable: Ready weights for supported models; in ComfyUI use ComfyUI-nunchaku.

  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

Requires the base model

akhaliq/Fibo-2.0-edit-IDM-2steps

This model is a LoRA / adapter. Download the base model first, then load this one on top.

  1. Fibo-2.0-devfamily root
  2. Fibo-2.0-edit-turbo
  3. Fibo-2.0-edit-IDM-2steps
  4. Fibo-2.0-edit-IDM-2steps-ComfyUI

New posts about versions of Fibo-2.0-dev (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 8All on Hugging Face
LoRAs and adapters 5All on Hugging Face

About the model

NidAll/Fibo-2.0-edit-IDM-2steps-ComfyUI
Author
NidAll
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments