CivArchive
    SD3 Medium Slightly Decensored (Perturbed) - %2 (clips+t5 fp16)
    NSFW
    Preview 15897786
    Preview 15897818
    Preview 15898204
    Preview 15898271
    Preview 15898346
    Preview 16010397
    Preview 16010441
    Preview 16010464
    Preview 16010498
    Preview 16010516
    Preview 16010583
    Preview 16010596
    Preview 16010597
    Preview 16010615
    Preview 16010757
    Preview 16010745
    Preview 16011130
    Preview 16011149
    Preview 16011154
    Preview 16011161

    credits to Thyri

    Perturbed Version of SD3 using script (replace input and output location)

    1. download sd3 from huggingface

    2. put script in a .py (replace input with sd3 model location and output location with where and what you want to name it). modify 0.01 to how much you want it to be perturbed (0.01 = 1%)

    3. execute the .py using "python your_py_script.py" (maybe you need to activate your virtual environment)


    import torch

    from safetensors.torch import load_file, save_file

    # Load the safetensors model

    model = load_file('__INPUT LOCATION__')

    # Extract the keys and tensors

    keys = model.keys()

    dic = {key: model[key] for key in keys}

    # Define parts to filter

    parts = ['diffusion_model']

    count = 0

    # Perturb the specified parts

    for k in keys:

    if all(part in k for part in parts):

    v = dic[k]

    print(f'{k}: {v.std()}')

    dic[k] += torch.normal(torch.zeros_like(v), v.std() * 0.01)

    count += 1

    print(f'Total perturbed tensors: {count}')

    # Save the perturbed model

    save_file(dic, '__OUTPUT LOCATION__')

    Description

    FAQ

    Comments (14)

    davchaJun 14, 2024· 8 reactions
    CivitAI

    I'm Thyri.

    oprahJun 14, 2024
    CivitAI

    Doesn't work at all for me. Got this error in ComfyUI:

    Error occurred when executing CLIPTextEncode: 'NoneType' object has no attribute 'tokenize' File "D:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 151, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 81, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 74, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\AI\ComfyUI_windows_portable\ComfyUI\nodes.py", line 57, in encode tokens = clip.tokenize(text) ^^^^^^^^^^^^^

    Queue size: 0

    Extra options

    Running

    Pending

    163192Jun 14, 2024· 3 reactions
    CivitAI

    It does seem to reduce body horror and improve the outputs in the limited testing I've done. Moving in the right direction!

    mizoku_azureJun 15, 2024
    CivitAI

    Won't load for me in Swarm UI.

    coyedi5985Jun 15, 2024· 9 reactions
    CivitAI

    I'm not sure if I want to both since the description is one long dick joke.

    4331997Jun 15, 2024
    CivitAI

    Do I need to execute that code? I just have the model itself loaded. If so I don't know how to run that code.

    hapemxgJun 15, 2024· 4 reactions
    CivitAI

    展示图片有点吓人😰

    bankenichiJun 15, 2024· 4 reactions
    CivitAI

    You really are on to something with the perturbed models, Im trying the 2% and it is doing much better than base SD3.

    Keep up the good work!

    HapseJun 15, 2024· 11 reactions
    CivitAI

    dic[k]

    Hehe

    androsynth7610Jun 16, 2024
    CivitAI

    Okay its 2B as well including clips and T5. So there is still a 4B and 8B one around. Its actually "small". Lots of strange chatter going on... i don't know to make heads or tails of it: https://www.reddit.com/r/StableDiffusion/comments/1dhd7vz/the_developer_of_comfy_who_also_helped_train_some/#lightbox I surely would love to get my hands on the 4B

    spammenotinoz190Jun 26, 2024
    CivitAI

    Thank-you, a big improvement, but still quite restrictive.

    JayNLJul 11, 2024
    CivitAI

    I have an unpublished adaptation of SD3 medium, nobody can see it right now, if you archive it, can you unarchive it when Civitai allows SD3 models? 🙏

    user1234123Jul 27, 2024· 1 reaction
    CivitAI

    Ah, now it's clear. A key part I was missing a long dic[k] += torch.normal(torch.zeros_like(v), v.std() * 0.01) it was all about me lacking the longer dic[k] part.

    Does v = dic[k] mean defining a micro_dic[k] part as larger than it is it could still work for this?

    wt_01333Sep 11, 2025
    CivitAI

    umm.. wow! Could you provide just script file for windows commant line or terminal, to activate it and make everything mathematically right?

    Checkpoint
    SD 3

    Details

    Downloads
    1,051
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    6/15/2024
    Updated
    4/27/2026
    Deleted
    4/27/2026

    Files

    sd3MediumSlightly_2ClipsT5Fp16.safetensors

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.