from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-8B")
model = PeftModel.from_pretrained(base_model, "navicore/fortean-qwen3-8b-advanced")
tokenizer = AutoTokenizer.from_pretrained("navicore/fortean-qwen3-8b-advanced")

prompt = "Question: What's your take on artificial intelligence?\n\nCharles Fort:"
# Generate response...

With Ollama

GGUF versions available at: navicore/fortean-qwen3-8b-advanced-GGUF

bash

# Download GGUF and Modelfile from the GGUF repo
ollama create fortean -f Modelfile
ollama run fortean "What do you think about UFOs?"

Training Details

Base model: Qwen3-8B (not instruct)
Method: LoRA fine-tuning
Training examples: ~200 high-quality Q&A pairs
Hardware: Apple M4 Pro

Example Output

Q: What's your view on cryptocurrency?

A: Digital gold that appears from nothing, vanishes into cryptographic ether, yet somehow purchases real coffee. I've documented seventeen mining facilities where the clocks run backward. In 1898, a banker in Prague reported similar temporal anomalies when counting certain denominations. The blockchain, they say, is immutable—yet I have records of transactions that preceded their own origins. Perhaps we're not mining coins but excavating something that was always there, waiting.