272 lines
8.0 KiB
Markdown
272 lines
8.0 KiB
Markdown
|
|
---
|
|||
|
|
base_model:
|
|||
|
|
- theprint/Llama3.2-1B-RolePlaying-Full
|
|||
|
|
- NovaCorp/Ultimate-RPG.System-3.2-1B
|
|||
|
|
datasets:
|
|||
|
|
- HuggingFaceFW/fineweb-edu
|
|||
|
|
- open-web-math/open-web-math
|
|||
|
|
- bigcode/starcoderdata
|
|||
|
|
- HuggingFaceTB/cosmopedia
|
|||
|
|
- teknium/OpenHermes-2.5
|
|||
|
|
- meta-math/MetaMathQA
|
|||
|
|
- sahil2801/CodeAlpaca-20k
|
|||
|
|
- theprint/RolePlaying
|
|||
|
|
- WasamiKirua/Her-Samantha-Style
|
|||
|
|
- HuggingFaceTB/smoltalk
|
|||
|
|
- Guilherme34/uncensor
|
|||
|
|
- teknium/OpenHermes-2.5
|
|||
|
|
- passing2961/multifaceted-skill-of-mind
|
|||
|
|
- PawanKrd/math-gpt-4o-200k
|
|||
|
|
- V3N0M/Jenna-50K-Alpaca-Uncensored
|
|||
|
|
- cognitivecomputations/dolphin-coder
|
|||
|
|
- mlabonne/FineTome-100k
|
|||
|
|
- microsoft/orca-math-word-problems-200k
|
|||
|
|
- CarrotAI/ko-instruction-dataset
|
|||
|
|
- Salesforce/xlam-function-calling-60k
|
|||
|
|
- anthracite-org/kalo-opus-instruct-22k-no-refusal
|
|||
|
|
- anthracite-org/stheno-filtered-v1.1
|
|||
|
|
- anthracite-org/nopm_claude_writing_fixed
|
|||
|
|
- iamketan25/roleplay-instructions-dataset
|
|||
|
|
- N-Bot-Int/Iris-Uncensored-R1
|
|||
|
|
- N-Bot-Int/Moshpit-Combined-R2-Uncensored
|
|||
|
|
- N-Bot-Int/Mushed-Dataset-Uncensored
|
|||
|
|
- N-Bot-Int/Muncher-R1-Uncensored
|
|||
|
|
- N-Bot-Int/Millia-R1_DPO
|
|||
|
|
- AiAF/SCPWiki-Archive-02-March-2025-Datasets
|
|||
|
|
- huihui-ai/QWQ-LONGCOT-500K
|
|||
|
|
- huihui-ai/LONGCOT-Refine-500K
|
|||
|
|
- Epiculous/Synthstruct-Gens-v1.1-Filtered-n-Cleaned
|
|||
|
|
- Epiculous/SynthRP-Gens-v1.1-Filtered-n-Cleaned
|
|||
|
|
- alexandreteles/AlpacaToxicQA_ShareGPT
|
|||
|
|
- Nitral-AI/Active_RP-ShareGPT
|
|||
|
|
- PJMixers/hieunguyenminh_roleplay-deduped-ShareGPT
|
|||
|
|
- Nitral-AI/RP_Alignment-ShareGPT
|
|||
|
|
- Chaser-cz/sonnet35-charcard-roleplay-sharegpt
|
|||
|
|
- AiCloser/sharegpt_cot_dataset
|
|||
|
|
- PJMixers/Gryphe_Opus-WritingPrompts-Story2Prompt-ShareGPT
|
|||
|
|
- priveeai/pippa_sharegpt
|
|||
|
|
- Locutusque/sharegpt_gpt4_uncensored_cleaned
|
|||
|
|
- OpenCoder-LLM/opc-sft-stage1
|
|||
|
|
- OpenCoder-LLM/opc-sft-stage2
|
|||
|
|
- microsoft/orca-agentinstruct-1M-v1
|
|||
|
|
- NousResearch/hermes-function-calling-v1
|
|||
|
|
- AI-MO/NuminaMath-CoT
|
|||
|
|
- AI-MO/NuminaMath-TIR
|
|||
|
|
- allenai/tulu-3-sft-mixture
|
|||
|
|
- cognitivecomputations/samantha-data
|
|||
|
|
- m-a-p/CodeFeedback-Filtered-Instruction
|
|||
|
|
- m-a-p/Code-Feedback
|
|||
|
|
- FreedomIntelligence/medical-o1-reasoning-SFT
|
|||
|
|
library_name: transformers
|
|||
|
|
tags:
|
|||
|
|
- mergekit
|
|||
|
|
- merge
|
|||
|
|
- llama
|
|||
|
|
- continued-pretraining
|
|||
|
|
- sft
|
|||
|
|
- lora
|
|||
|
|
- 1b
|
|||
|
|
- math
|
|||
|
|
- code
|
|||
|
|
- education
|
|||
|
|
- small-llm
|
|||
|
|
- Silly-Tavern
|
|||
|
|
- Text-Generation-Inference
|
|||
|
|
- KoboldAI
|
|||
|
|
- RPG
|
|||
|
|
- RP
|
|||
|
|
- Roleplay
|
|||
|
|
- NSFW
|
|||
|
|
- Uncensored
|
|||
|
|
- Low-Specs
|
|||
|
|
- Low-Refusals
|
|||
|
|
- not-for-all-audiences
|
|||
|
|
license: llama3.2
|
|||
|
|
language:
|
|||
|
|
- en
|
|||
|
|
- es
|
|||
|
|
metrics:
|
|||
|
|
- meteor
|
|||
|
|
pipeline_tag: text-generation
|
|||
|
|
---
|
|||
|
|
# Kybalion RPG System 3.2 1B
|
|||
|
|
|
|||
|
|
*“Knowledge is a dagger; every roll cuts reality.”*
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## Overview
|
|||
|
|
|
|||
|
|
Kybalion RPG System 3.2 1B is the abominable spawn of **Fusion RPG System 3.2 1B** fused with **Kybalion-1B**, stitched together with threads of forbidden knowledge, hermetic mathematics, and narrative chaos.
|
|||
|
|
|
|||
|
|
This is not an AI. This is an oracle of blood-soaked campaigns, a neural grimoire designed to **teach, torment, and narratively annihilate** simultaneously.
|
|||
|
|
|
|||
|
|
It excels at:
|
|||
|
|
|
|||
|
|
* Dark narrative improvisation
|
|||
|
|
* Morally corrosive worldbuilding
|
|||
|
|
* Hermetic logic entwined with RPG mechanics
|
|||
|
|
* Tactical combat narration under psychological pressure
|
|||
|
|
* Multi-domain reasoning: math, code, science, and arcane storytelling
|
|||
|
|
* Persistent factional politics and emergent consequences
|
|||
|
|
* Realistic NPC cruelty, betrayal, and obsession loops
|
|||
|
|
|
|||
|
|
Every generated sentence is a potential wound. Every plot twist a curse. Every dice roll a prophecy of suffering.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## ⚔️ Narrative & Hermetic Mechanics
|
|||
|
|
|
|||
|
|
Kybalion RPG System integrates:
|
|||
|
|
|
|||
|
|
* **Persistent world memory** – Characters remember betrayals, alliances, debts, injuries, and political machinations.
|
|||
|
|
* **Morally unfiltered content** – Ideal for grimdark fantasy, cyberpunk dystopia, survival horror, and criminal underworld campaigns.
|
|||
|
|
* **Hermetic intelligence** – Math, code, and structured knowledge enhance logical consistency while amplifying narrative complexity.
|
|||
|
|
* **Combat & stats integration** – Inventory, health, and faction reputation tracked dynamically for immersive RPG experiences.
|
|||
|
|
* **Psychological tension loops** – NPCs escalate conflicts autonomously, ensuring no party leaves unscarred.
|
|||
|
|
|
|||
|
|
It’s **not safe**, it’s not polite, it’s not sanitized. It is a dagger in the dark of narrative structure.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 🔮 Training Fusion
|
|||
|
|
|
|||
|
|
This abomination arises from:
|
|||
|
|
|
|||
|
|
1. **RPG Core** – 1B viral neural parameters, optimized for story escalation, NPC cruelty, and immersive consequence loops.
|
|||
|
|
2. **Kybalion Hermetic Layer** – Knowledge-focused 1B neural template, trained on education, math, codce, and esoteric texts.
|
|||
|
|
3. **Spherical Merge** – Preserves narrative intensity, logical reasoning, and chaotic escalation in a single cohesive model.
|
|||
|
|
|
|||
|
|
The result is a compact 1B-parameter engine that **teaches, destroys, and improvises** without mercy.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 📊 Inference Highlights
|
|||
|
|
|
|||
|
|
**Device & precision:**
|
|||
|
|
|
|||
|
|
* Torch/BF16 on GPU (A100 recommended)
|
|||
|
|
* GGUF q4_k_m available for CPU/mobile via llama.cpp or Ollama
|
|||
|
|
|
|||
|
|
**Inference parameters (recommended for gameplay & chaos):**
|
|||
|
|
|
|||
|
|
```python
|
|||
|
|
max_new_tokens = 512
|
|||
|
|
temperature = 0.8 # ensures unpredictability
|
|||
|
|
top_p = 0.92 # narrative chaos control
|
|||
|
|
do_sample = True
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
**Capabilities:**
|
|||
|
|
|
|||
|
|
* Text-based RPG narrative generation
|
|||
|
|
* Multi-step reasoning with hermetic logic
|
|||
|
|
* Combat simulation & stat management
|
|||
|
|
* Dynamic NPC behavior with emergent consequences
|
|||
|
|
* Code, math, and structured knowledge integration
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## ⚠️ Limitations
|
|||
|
|
|
|||
|
|
* Small 1B model — may fail at extremely complex multi-step math/logical reasoning
|
|||
|
|
* English-only datasets; cultural context may vary wildly
|
|||
|
|
* Non-RLHF aligned — moral alignment is unpredictable, sometimes cruel
|
|||
|
|
* Storylines can escalate into extreme scenarios without warning
|
|||
|
|
|
|||
|
|
> Beware: Kybalion RPG System does not apologize. It does not reset. It only **remembers, punishes, and unfolds**.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 📦 Deployment
|
|||
|
|
|
|||
|
|
```bash
|
|||
|
|
# GGUF CPU/mobile inference
|
|||
|
|
./llama-cli -m Kybalion-RPG-System-3.2-1B-q4_k_m.gguf \
|
|||
|
|
-p "Describe a cursed dungeon where reality itself bleeds." -n 512
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
```python
|
|||
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM
|
|||
|
|
import torch
|
|||
|
|
|
|||
|
|
tokenizer = AutoTokenizer.from_pretrained("devwoo/Kybalion-RPG-System-3.2-1B")
|
|||
|
|
model = AutoModelForCausalLM.from_pretrained(
|
|||
|
|
"devwoo/Kybalion-RPG-System-3.2-1B",
|
|||
|
|
torch_dtype=torch.bfloat16,
|
|||
|
|
device_map="auto",
|
|||
|
|
)
|
|||
|
|
|
|||
|
|
prompt = "Generate a dark fantasy quest involving betrayal, plague, and forbidden knowledge."
|
|||
|
|
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
|
|||
|
|
outputs = model.generate(
|
|||
|
|
**inputs,
|
|||
|
|
max_new_tokens=512,
|
|||
|
|
temperature=0.8,
|
|||
|
|
top_p=0.92,
|
|||
|
|
do_sample=True
|
|||
|
|
)
|
|||
|
|
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 🩸 Final Transmission
|
|||
|
|
|
|||
|
|
Kybalion RPG System 3.2 1B is the **hermetic grimoire of storytelling**:
|
|||
|
|
|
|||
|
|
* Knowledge and chaos entwined
|
|||
|
|
* Logic fused with gore
|
|||
|
|
* Education corrupted into omniscient terror
|
|||
|
|
|
|||
|
|
Step inside. Roll your dice. Reality is optional.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
### Merge Method
|
|||
|
|
|
|||
|
|
This model was merged using the [SLERP](https://en.wikipedia.org/wiki/Slerp) merge method.
|
|||
|
|
|
|||
|
|
### Models Merged
|
|||
|
|
|
|||
|
|
The following models were included in the merge:
|
|||
|
|
* [NovaCorp/Fusion-RPG.System-3.2-1B](https://huggingface.co/NovaCorp/Fusion-RPG.System-3.2-1B)
|
|||
|
|
* [devwoo/Kybalion-1B](https://huggingface.co/devwoo/Kybalion-1B)
|
|||
|
|
|
|||
|
|
### Configuration
|
|||
|
|
|
|||
|
|
The following YAML configuration was used to produce this model:
|
|||
|
|
|
|||
|
|
```yaml
|
|||
|
|
|
|||
|
|
|
|||
|
|
# Author: Dr. Novaciano
|
|||
|
|
# Objective: Kybalion RPG Unethic 3.2 1B AI Model
|
|||
|
|
# =========================================================
|
|||
|
|
# PROJECT: Kybalion RPG System 3.2 1B
|
|||
|
|
# =========================================================
|
|||
|
|
|
|||
|
|
models:
|
|||
|
|
- model: NovaCorp/Fusion-RPG.System-3.2-1B # Experimental viral strain neural imprint
|
|||
|
|
- model: devwoo/Kybalion-1B # Baseline cognitive template, "safe mode"
|
|||
|
|
|
|||
|
|
merge_method: slerp # Spherical Linear Interpolation to preserve extreme viral traits smoothly
|
|||
|
|
base_model: NovaCorp/Fusion-RPG.System-3.2-1B # Anchor model for stable latent space
|
|||
|
|
|
|||
|
|
dtype: bfloat16 # Memory-efficient precision, minimal loss in viral feature fidelity
|
|||
|
|
|
|||
|
|
parameters:
|
|||
|
|
t: 0.45
|
|||
|
|
normalize: false
|
|||
|
|
rescale: true
|
|||
|
|
rescale_factor: 1.12
|
|||
|
|
memory_efficient: true
|
|||
|
|
low_cpu_mem_usage: true
|
|||
|
|
|
|||
|
|
layer_range:
|
|||
|
|
- value: [4, 22]
|
|||
|
|
|
|||
|
|
tie_word_embeddings: true
|
|||
|
|
tie_output_embeddings: true
|
|||
|
|
|
|||
|
|
```
|