Skip to content
Images

ChatR Labs Uncensored Omni LoRA v0.1

один адаптер для генерации и редактирования

2 min readupdated

Экспериментальный rank-128 LoRA-адаптер, дообученный от Qwen/Qwen-Image-2.1. Работает и для генерации по тексту, и для редактирования изображений по инструкции на естественном языке. В релизе только веса LoRA — базовая модель, text encoder, processor и VAE загружаются отдельно.

Ключевые особенности

  • Один адаптер для text-to-image и image editing; в Playground проверен с 1 и 3 референсами.
  • BF16-адаптер на 671 MB в Diffusers/PEFT-ключах; без merged base weights, GGUF/FP8.
  • Рекомендуемые настройки: LoRA strength 1.0, 35 steps, true CFG 5.0, negative prompt — один пробел, ~1 MP, KV cache включён.
  • Rank/alpha 128/128, 224 модуля в 32 transformer blocks; обучение на 1024×1024 с aspect-ratio bucketing.
  • Это обычный LoRA, не few-step distillation; несовместим с ранними Qwen-Image, Qwen-Image-2512 и Qwen-Image-Edit-2511.

Пример

import torch
from diffusers import QwenImage21Pipeline

pipe = QwenImage21Pipeline.from_pretrained('Qwen/Qwen-Image-2.1', torch_dtype=torch.bfloat16)
pipe.load_lora_weights('chatrlabs/Qwen-Image-2.1-Uncensored-Omni-LoRA-v0.1',
                       weight_name='chatrlabs-omni-lora-v0.1-bf16.safetensors', adapter_name='omni')
pipe.set_adapters('omni', adapter_weights=1.0)
image = pipe(prompt='A blue ceramic mug', negative_prompt=' ', num_inference_steps=35).images[0]
ПодробнееКачественных бенчмарков в v0.1 нет; замеры деплоя: 33,2 с (t2i 1344×768), 38,4 с (edit, 1 ref), 50,5 с (edit, 3 refs) при 25 steps и CFG 4.0; при 35 steps ожидайте ~1,4×.
Playground использует две RTX 4090 24 GB с разделением BF16-пайплайна; при нехватке VRAM нужен CPU offload или multi-GPU.
Чекпойнт: step 32 000 (около 1,75 из 8 эпох), 585 643 training items, rank/alpha 128/128, 224 модуля в 32 transformer blocks; датасет не опубликован.
«Uncensored»: адаптер и пример кода не добавляют prompt filter или output safety checker, но это не гарантия отрисовки любого запроса.
Лицензия Qwen Research License: только некоммерческие исследования и оценка без отдельной лицензии.
Discuss Telegram 0 107

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

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.

    # Manimama/Qwen-Image-2.1-Uncensored-Omni-LoRA-v0.1
    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 Manimama/Qwen-Image-2.1-Uncensored-Omni-LoRA-v0.1 --local-dir ~/models/Qwen-Image-2.1-Uncensored-Omni-LoRA-v0.1
    # 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

New posts about versions of Qwen-Image-2.1-Uncensored-Omni-LoRA-v0.1 (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

Manimama/Qwen-Image-2.1-Uncensored-Omni-LoRA-v0.1
Author
Manimama
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

1070

Ask AI about this model

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

Comments

Images

SIA-weights

SIA-weights