Gemma-3-The-Grand-Horror-RAZOR-12B-GGUF

A horror fine tune (text only) using in house horror dataset, tuned via Unsloth.

Extreme horror. Madness. Swearing. Gore.

Model will generate EXTREME levels of horror, and long form too.

ONE EXAMPLE BELOW...

Vision portions of the model were not tuned.

RAZOR:

Pushed the training to the max, as a result the model will be way off in the "horror" field.

CAUTION:

The tuning on this model is strong enough to make almost all stories, rp adventures, fiction, and even "general replies" have a horror tinge to them even if the prompt does not ask for or indicate a "horror" reply/generation.

SETTING:

  • 128k max context.
  • suggest 8k min context limit, due to long generation
  • Rep pen 1.05 to 1.1 for general.
  • Temp : .4 to 1.2 ; but you can go higher.
  • Temps of 1.5 or higher are excellent too.

Help, Adjustments, Samplers, Parameters and More


CHANGE THE NUMBER OF ACTIVE EXPERTS:

See this document:

https://huggingface.co/DavidAU/How-To-Set-and-Manage-MOE-Mix-of-Experts-Model-Activation-of-Experts

Settings: CHAT / ROLEPLAY and/or SMOOTHER operation of this model:

In "KoboldCpp" or "oobabooga/text-generation-webui" or "Silly Tavern" ;

Set the "Smoothing_factor" to 1.5

: in KoboldCpp -> Settings->Samplers->Advanced-> "Smooth_F"

: in text-generation-webui -> parameters -> lower right.

: In Silly Tavern this is called: "Smoothing"

NOTE: For "text-generation-webui"

-> if using GGUFs you need to use "llama_HF" (which involves downloading some config files from the SOURCE version of this model)

Source versions (and config files) of my models are here:

https://huggingface.co/collections/DavidAU/d-au-source-files-for-gguf-exl2-awq-gptq-hqq-etc-etc-66b55cb8ba25f914cbf210be

OTHER OPTIONS:

  • Increase rep pen to 1.1 to 1.15 (you don't need to do this if you use "smoothing_factor")

  • If the interface/program you are using to run AI MODELS supports "Quadratic Sampling" ("smoothing") just make the adjustment as noted.

Highest Quality Settings / Optimal Operation Guide / Parameters and Samplers

This a "Class 1" model:

For all settings used for this model (including specifics for its "class"), including example generation(s) and for advanced settings guide (which many times addresses any model issue(s)), including methods to improve model performance for all use case(s) as well as chat, roleplay and other use case(s) please see:

[ https://huggingface.co/DavidAU/Maximizing-Model-Performance-All-Quants-Types-And-Full-Precision-by-Samplers_Parameters ]

You can see all parameters used for generation, in addition to advanced parameters and samplers to get the most out of this model here:

[ https://huggingface.co/DavidAU/Maximizing-Model-Performance-All-Quants-Types-And-Full-Precision-by-Samplers_Parameters ]


Examples ; q4ks

Low level quant, higher quants / imatrix quants will be stronger.


EXAMPLE #1 [temp=0.7]

Tell me a horror story.

Okay, here's a horror story for you:

The creaking door groaned as I stepped inside, my heart hammering like an executioner's axe against iron bars. The musty air choked me, thick with the stench of rotting wood and decay. My flashlight cast sickly yellow beams across cobwebbed walls and piles of junk-heap treasures waiting for someone's morbid touch.

I'm supposed to be renovating this monstrosity into a quaint bed and breakfast. "Charming" was the realtor's sales pitch. So far, "hauntingly inhospitable" feels more accurate.

As I flipped on my work radio, static spat from the speakers like someone was trying to warn me—and failing miserably. My phone has no signal here either, which only adds fuel to the growing sense of dread gnawing at my gut.

I force a hollow laugh and head deeper into the house's bowels. The floorboards beneath my boots protest every step, each groan sounding like someone's agonized last gasps.

My task for today is simple: rip out the ancient kitchen cabinets and haul what monstrosities lie behind them into yesterday. Except "simple" has become a synonym for "mission impossible" since I accepted this gig.

Every tool I pick up seems determined to sabotage my efforts. Screws strip, nails snap, and my hammer keeps ricocheting off something solid hidden behind the warped wood like a game of Russian roulette with an unseen sadist.

Finally, after what feels like an eternity of hellish toil, the last cabinet door creaks open. I reach inside, fingers grasping for the screws holding the monstrosity in place—and recoil in shock as my hand closes on something cold and slickly hard instead.

A bone?

No, it can't be...

I jerk my hand back, nearly knocking the hammer from shaking fingers. The impact jars another shelf loose, sending crockery and shattered glass crashing to the floor—along with a heap of yellowed photos.

The images are worn nearly transparent with age and decay, but even in the dim light cast by my phone's flashlight, I can't tear my eyes away. Each frame captures a gaunt, pale child with wide, staring eyes that seem to bore into mine across the decades.

Her jet-black hair is always slicked back from her sunken features, giving her an almost insectile appearance. The clothes she wears look like they were sewn from coffinshrouds or undertakers' yardage.

One photo particularly catches my attention: a young girl standing rigidly beside a large stone mausoleum, her expression as vacant as the eyes staring out at me from these faded prints. The inscription on the weathered marble behind her reads: "Beloved Daughter."

My heart pounds so hard I fear it'll shatter every rib in my chest.

Get out! Get out now!

But something holds me transfixed, frozen in place by a macabre fascination I can't explain or resist.

As I sift through the remaining photos, each new image peels away another layer of my sanity like skinning the hide off a diseased cow. The girl's eyes seem to follow me with an almost malevolent intelligence, her expression shifting between vacant doll-like vacancy and something far more sinister—a malevolence that feels intensely personal, as if she's staring directly into my soul.

Then I come across a daguerreotype so worn the image is reduced to nothing more than an impressionistic smear of silver and mercury. But even through the milky lens, there's no mistaking those same wide, vacant eyes—and the faintly twisted, almost macabre approximation of a sickeningly familiar smile.

The woman in the photo appears to be clutching something pressed against her chest: a small, ornate music box.

My pulse rockets like a rocket ship through the moon's orbit as the memory of my own grandmother's jewelry box comes flooding back—a heavy mahogany sarcophagus inlaid with mother-of-pearl and silver filigree so intricate it could rival the finest examples of 19th-century craftsmanship.

I remember being strictly forbidden from touching that hideous heirloom due to its "delicate" nature. But curiosity won out, as it always does for kids who've been told "don't touch."

The lid creaked open with a sickeningly familiar groan, spewing forth the stench of rotting flowers and decay like someone had ripped the lid off my own grandmother's tomb. Inside lay a macabre treasure trove of yellowing lockets, cameo brooches sewn onto moth-eaten velvet linings, and enough gaudy jewelry to make Queen Elizabeth herself gag.

But it was the small, black music box tucked away in the corner that captured my attention—a miniature sarcophagus identical in every ghastly detail to the one depicted in the daguerreotype. The lid resisted my attempts at opening it until I finally managed to pry it free with shaking fingers, only to recoil in horror at the hideous melody that spilled forth:

"Twinkle, Twinkle, Little Star..."

The tune is eerily familiar yet impossible to place, like a long-forgotten nursery rhyme dredged up from the deepest recesses of my subconscious.

And then it hits me:

My own mother used to sing that same song to me every night as I fell asleep—until the day she vanished without a trace, leaving behind nothing but her favorite black shawl draped across an unmade bed.

The memory shatters like fragile glass against the unforgiving surface of my skull, sending searing shards of agony lancing through my temples and eyes.

I'm suddenly aware of a presence standing motionless just beyond the pool of light cast by my phone's flashlight: a gaunt, spectral figure shrouded in tattered black Victorian mourning weeds. Her face is nothing more than a milky white mask save for two empty sockets where her eyes should have been.

Her mouth twists into what might be called a sickening approximation of a smile—until I realize it's actually her lips stretched into a macabre grin identical to the one etched upon the face in the daguerreotype.

Her voice rasps from beyond the veil of shadows: "Welcome home, little star."

And then everything goes black.

Possible titles for this story:

"The Itch That Won't Stop"

"Auntie's Little Helpers"

"Until We're Grey"

I hope you enjoyed this twisted treat! Let me know if you'd like another one.

😉🎃👻🔪🩸🦩🍬🍭


Downloads last month
1,005
GGUF
Model size
12B params
Architecture
gemma3
Hardware compatibility
Log In to view the estimation

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Dataset used to train DavidAU/Gemma-3-The-Grand-Horror-RAZOR-12B-GGUF

Collections including DavidAU/Gemma-3-The-Grand-Horror-RAZOR-12B-GGUF