navicore/fortean-qwen3-8b-advanced-GGUF

Fortean-Qwen3-8B GGUF

This is a GGUF quantized version of the Fortean-Qwen3-8B model, fine-tuned to emulate Charles Fort's distinctive writing style about anomalous phenomena.

Model Details

• Base Model: Qwen3-8B
• Fine-tuning: LoRA adapter trained on Charles Fort's complete works
• Quantization: Q4_K_M (4-bit quantization, ~4.9GB)
• Use Case: Creative writing, anomaly analysis, historical curiosity exploration

Installation

1. Install Ollama: https://ollama.ai
2. Create the model:

ollama create fortean -f Modelfile.final

3. Run the model:

ollama run fortean

Example Usage

User: What are your thoughts on the Bermuda Triangle?

Charles Fort: The Bermuda Triangle, that persistent enigma that haunts our collective imagination, presents patterns that would alarm those who prefer their reality comfortably predictable. I've documented instances where similar manifestations occur across multiple continents with a periodicity that transcends local explanation...

Model Characteristics

• Generates responses in Charles Fort's distinctive style
• References obscure 19th century sources
• Connects seemingly unrelated phenomena
• Exhibits dry wit and skepticism of orthodox explanations
• Stops generating at reasonable lengths

Optimal Usage

Prompt Format

For best results, use this prompt format:

Question: [your question here]

Charles Fort:

Recommended Parameters

• Temperature: 0.7-0.8
• Top P: 0.85-0.9
• Repeat Penalty: 1.1-1.2
• Max Tokens: 250-300

Stop Tokens

Configure these stop tokens to prevent multiple responses:

• <|endoftext|>
• <|im_end|>
• \nQuestion:
• \n\nQuestion:

Example Ollama Modelfile

FROM fortean-q4_k_m.gguf

TEMPLATE """Question: {{ .Prompt }}

Charles Fort: {{ .Response }}"""

PARAMETER temperature 0.8
PARAMETER top_p 0.9
PARAMETER repeat_penalty 1.1
PARAMETER num_predict 300
PARAMETER stop "<|endoftext|>"
PARAMETER stop "<|im_end|>"
PARAMETER stop "\nQuestion:"
PARAMETER stop "\n\nQuestion:"

Usage with llama.cpp

./main -m fortean-q4_k_m.gguf -p "Question: What are your thoughts on red rain?\n\nCharles Fort:" -n 300 --temp 0.8

Usage with Python (llama-cpp-python)

from llama_cpp import Llama

llm = Llama(model_path="fortean-q4_k_m.gguf", n_ctx=2048)

prompt = "Question: What are your thoughts on red rain?\n\nCharles Fort:"
response = llm(prompt, max_tokens=300, temperature=0.8, stop=["<|endoftext|>", "\nQuestion:"])
print(response['choices'][0]['text'])

Known Issues

• Without proper stop tokens, the model may generate multiple responses
• Best results with clear, direct questions about anomalous phenomena

Credits

Fine-tuned by navicore using Charles Fort's complete works from Project Gutenberg.