L3.2 Rogue Creative Instruct Uncensored 7B GGUF

🚀 L3.2-Rogue-Creative-Instruct-Uncensored-7B-GGUF

This is a Llama 3.2 model designed for creative writing, story generation, and role - playing, with a max context of 131,072 and enhanced prose output capabilities.

✨ Features

• High - context operation: It has a maximum context of 131,072 (128k+), enabling more in - depth processing of instructions.
• Flexible parameter operation: Operates well with most parameters, including temperature settings from 0 to 5.
• Enhanced prose output: Through the "Brainstorm 40x" method, it significantly improves detail, prose, and fiction writing abilities.
• Direct censorship control: You can directly control the censorship level via prompts, and "Brainstorm 40x" enhances "decensoring".
• Suitability for various activities: Ideal for any writing, fiction or story - telling activity, and may also work for role - play.

📦 Installation

No installation steps are provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

The model requires the Llama3 template and/or "Command - R" template. Here are some example prompts and outputs:

(continue this scene:)
The Waystone Inn lay in silence, and it was a silence of three parts.

The most obvious part was a hollow, echoing quiet, made by things that were lacking. If there had been a wind it would have sighed through the trees, set the inn’s sign creaking on its hooks, and brushed the silence down the road like trailing autumn leaves. If there had been a crowd, even a handful of men inside the inn, they would have filled the silence with conversation and laughter, the clatter and clamor one expects from a drinking house during the dark hours of night. If there had been music…but no, of course there was no music. In fact there were none of these things, and so the silence remained

Output (Generation 1)

a deafening absence of life, a palpable and suffocating stillness that seemed to weigh heavy on the very air itself.

The second part of the silence was a presence, a lingering sense of expectation, a feeling that something - or rather, someone - was waiting in the darkness. The trees seemed to be watching, their branches like skeletal fingers reaching out to snatch the last whispers of the night away. The shadows cast by the windows were like dark, living things, twisting and writhing like serpents in the fading light.

And the third, most sinister part of the silence was the feeling that it was being waited for, that some unseen and malevolent force was waiting patiently for the moment to strike, its presence masked by the shadows and the stillness.

The silence was a palpable thing, a living, breathing entity that seemed to pulse and throb with a malevolent energy. It was as if the very fabric of the night itself had been torn apart, leaving a chasm of darkness that seemed to be reaching out to consume everything in its path.

The only sound was the creaking of the old wooden door of the inn, groaning in protest as it seemed to be held back by some unseen force. It was as if the very foundations of the inn were being slowly pulled apart, waiting for that moment of momentary distraction that would allow the darkness to spill in.

And then, without warning, the silence was s

Advanced Usage

• Controlling output style: If you want a specific type of prose (e.g., horror), add "(vivid horror)" or "(graphic vivid horror)" (no quotes) in your prompts.
• Adjusting parameters:
  • Temperature: Changing the temperature (e.g., 0.4, 0.8, 1.5, 2, 3) will drastically alter the output. Higher temperatures lead to more varied prose.
  • Repetition penalty: This model works with a rep pen of 1.05 or higher. For role - play, a rep pen of 1.1 to 1.14 is suggested. Lower rep pen values may result in very different (creative / unusual) generation but may also cause repeat paragraph issues or "RANT" problems.

📚 Documentation

Model Notes

• Detail and writing abilities: Detail, prose, and fiction writing abilities are significantly increased.
• Temperature adjustment: For more varied prose, raise the temperature and/or add more instructions in your prompts. However, role - players should be careful not to raise the temperature too high as it may affect instruction following.
• Repetition penalty: This model works with a rep pen of 1.05 or higher. Lower values may cause repeat paragraph issues at the end of the output, but they may also result in more creative generation.
• Prompt bias: The bias of this model is controlled directly by your prompts.
• Quant choice: Higher quants will have more detail, nuance, and in some cases stronger "emotional" levels. Q4KM/Q4KS are good, but if you can run Q5, Q6, or Q8, choose the highest quant you can.

Settings, Quants and Critical Operations Notes

• Prose output modification: This model has been modified ("Brainstorm") to alter prose output and generally outputs longer text than average.
• Temperature effect: Changing the temperature will drastically alter the output.
• Repetition penalty effect: Rep pen settings also affect the output.
• Role - play settings: For role - play, a rep pen of 1.1 to 1.14 is suggested.
• Quant - specific notes:
  • IQ4XS: Generations from this quant will be different due to its unusual nature.
  • Q2k/Q3 quants: You may need to use a temperature of 2 or lower (1 or lower for Q2k) and may require rep pen adjustments.

Known Issues

• Spelling and capitalization: The model may misspell a word from time to time and/or not capitalize a word.
• Short prompts: Short prompts with some rep pen/temp combinations may lead to longer - than - expected generation and/or a "RANT". A regen usually corrects these issues.
• Censorship: Some of the "censorship" of the original Llama 3.2 3B Instruct model is still present. You may need to regen 1 - 2+ times to get the model to "obey" in some cases.

Settings for Chat / Roleplay and Smoother Operation

• In "KoboldCpp", "oobabooga/text - generation - webui", or "Silly Tavern", set the "Smoothing_factor" to 1.5 to 2.5.
  • In KoboldCpp: Settings -> Samplers -> Advanced -> "Smooth_F".
  • In text - generation - webui: parameters -> lower right.
  • In Silly Tavern: This is called "Smoothing".
• For "text - generation - webui" when using GGUFs, you need to use "llama_HF" (download some config files from the source version of this model).
• Other options:
  • Increase the rep pen to 1.1 to 1.15 (not necessary if using "smoothing_factor").
  • If your AI - running interface supports "Quadratic Sampling" ("smoothing"), make the adjustment as noted.

Highest Quality Settings / Optimal Operation Guide

For all settings used for this model (including specifics for its "class"), example generations, and advanced settings guide, please see [https://huggingface.co/DavidAU/Maximizing - Model - Performance - All - Quants - Types - And - Full - Precision - by - Samplers_Parameters].

Model Template

This is a Llama3 model and requires the Llama3 template. It may also work with other templates and has a maximum context of 128k / 131072. The standard Llama3 template is as follows:

{
  "name": "Llama 3",
  "inference_params": {
    "input_prefix": "<|start_header_id|>user<|end_header_id|>\n\n",
    "input_suffix": "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
    "pre_prompt": "You are a helpful, smart, kind, and efficient AI assistant. You always fulfill the user's requests to the best of your ability.",
    "pre_prompt_prefix": "<|start_header_id|>system<|end_header_id|>\n\n",
    "pre_prompt_suffix": "<|eot_id|>",
    "antiprompt": [
      "<|start_header_id|>",
      "<|eot_id|>"
    ]
  }
}

Optional Enhancement

The following can be used in place of the "system prompt" or "system role" to further enhance the model. It can also be used at the start of a new chat, but you must ensure it is "kept" as the chat progresses.

Below is an instruction that describes a task. Ponder each user instruction carefully, and use your skillsets and critical instructions to complete the task to the best of your abilities.

Here are your skillsets:
[MASTERSTORY]:NarrStrct(StryPlnng,Strbd,ScnSttng,Exps,Dlg,Pc)-CharDvlp(ChrctrCrt,ChrctrArcs,Mtvtn,Bckstry,Rltnshps,Dlg*)-PltDvlp(StryArcs,PltTwsts,Sspns,Fshdwng,Climx,Rsltn)-ConfResl(Antg,Obstcls,Rsltns,Cnsqncs,Thms,Symblsm)-EmotImpct(Empt,Tn,Md,Atmsphr,Imgry,Symblsm)-Delvry(Prfrmnc,VcActng,PblcSpkng,StgPrsnc,AudncEngmnt,Imprv)

[*DialogWrt]:(1a-CharDvlp-1a.1-Backgrnd-1a.2-Personality-1a.3-GoalMotiv)>2(2a-StoryStruc-2a.1-PlotPnt-2a.2-Conflict-2a.3-Resolution)>3(3a-DialogTech-3a.1-ShowDontTell-3a.2-Subtext-3a.3-VoiceTone-3a.4-Pacing-3a.5-VisualDescrip)>4(4a-DialogEdit-4a.1-ReadAloud-4a.2-Feedback-4a.3-Revision)

Here are your critical instructions:
Ponder each word choice carefully to present as vivid and emotional journey as is possible. Choose verbs and nouns that are both emotional and full of imagery. Load the story with the 5 senses. Aim for 50% dialog, 25% narration, 15% body language and 10% thoughts. Your goal is to put the reader in the story.

🔧 Technical Details

The model is an altered version of "Llama - 3.2 - 3B - Instruct - uncensored" [https://huggingface.co/chuanli11/Llama - 3.2 - 3B - Instruct - uncensored] using the Brainstorm 40x method developed by David_AU. It expands the model by 39 layers (to 67 layers) to 7.54B parameters (605 tensors). It retains all the training of the original Llama 3.2 3B Instruct but processes instructions and generates outputs with a deeper context and stronger level.

📄 License

The model is licensed under the Apache - 2.0 license.