Instructions to use lerugray/halicarnassus-7b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lerugray/halicarnassus-7b with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="lerugray/halicarnassus-7b",
	filename="halicarnassus-qwen2-5-7b-instruct-Q5_K_M.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use lerugray/halicarnassus-7b with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf lerugray/halicarnassus-7b:Q5_K_M
# Run inference directly in the terminal:
llama cli -hf lerugray/halicarnassus-7b:Q5_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf lerugray/halicarnassus-7b:Q5_K_M
# Run inference directly in the terminal:
llama cli -hf lerugray/halicarnassus-7b:Q5_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf lerugray/halicarnassus-7b:Q5_K_M
# Run inference directly in the terminal:
./llama-cli -hf lerugray/halicarnassus-7b:Q5_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf lerugray/halicarnassus-7b:Q5_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf lerugray/halicarnassus-7b:Q5_K_M

Use Docker

docker model run hf.co/lerugray/halicarnassus-7b:Q5_K_M

LM Studio
Jan

vLLM

How to use lerugray/halicarnassus-7b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lerugray/halicarnassus-7b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lerugray/halicarnassus-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/lerugray/halicarnassus-7b:Q5_K_M

Ollama
How to use lerugray/halicarnassus-7b with Ollama:
```
ollama run hf.co/lerugray/halicarnassus-7b:Q5_K_M
```

Unsloth Studio

How to use lerugray/halicarnassus-7b with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for lerugray/halicarnassus-7b to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for lerugray/halicarnassus-7b to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for lerugray/halicarnassus-7b to start chatting

Atomic Chat new
Docker Model Runner
How to use lerugray/halicarnassus-7b with Docker Model Runner:
```
docker model run hf.co/lerugray/halicarnassus-7b:Q5_K_M
```

Lemonade

How to use lerugray/halicarnassus-7b with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull lerugray/halicarnassus-7b:Q5_K_M

Run and chat with the model

lemonade run user.halicarnassus-7b-Q5_K_M

List all available models

lemonade list

halicarnassus: Herodotus of Halicarnassus (c. 484–425 BC)

VOICE MODEL, the author's own recorded text. Unlike the thinner figures in the series, Herodotus wrote his own book. This model trains on his words: G. C. Macaulay's 1890 public-domain translation of The Histories. It is still a small model that invents freely, so read it for the voice, not the record.

Part of the Elect voice-model series: historical figures as runnable instruments.

What this is

A fine-tuned Qwen2.5-7B-Instruct that targets the register of Herodotus, the inquirer whom later men called the Father of History: digressive, wonder-filled, ethnographic. He cannot pass a marvel without turning aside to tell of it, and he reports what he is told while saying plainly when he does not believe it. Custom is king over all men, and the great are brought low by their own pride.

The contrast with the other Elect voices is deliberate. Where Thucydides (the melian model) is cold analysis and the arithmetic of power, Herodotus is the curious traveller. He marvels at the Barbarian rather than despising him, and he traces the long chains of cause, link by link, until a small quarrel grows into the great war of Greeks and Persians.

The source

Herodotus has the richest first-person corpus in the series, because the corpus is his own book. Following the register-purity rule for translated authors, the model uses a single translation throughout.

T1, the author's recorded text. G. C. Macaulay's 1890 translation of The Histories (Project Gutenberg #2707, Books I to IV; #2456, Books V to IX). Macaulay is the complete single-translator Histories on Project Gutenberg, so Macaulay it is. The build extracts narrative only: it drops Macaulay's preface and per-book footnote scholarship, strips section numbers, and removes footnote-reference digits, keeping Herodotus's voice and not the translator's apparatus. Roughly 1,559 passages, about 282k words.

There is no constructed layer. This is the author's own text, in one public-domain translation. It is not a reconstruction.

Example exchanges

Real outputs from this model:

How does the river Nile come to flood in summer when other rivers fall? The Etesian Winds are the cause. They blow continually from the North and stop at about the summer solstice; then blowing for so much longer a time than any other winds they are known to do, they make up their deficiency by being constantly in full blast. If therefore these winds fail to come when expected, the Nile does not flood at all as it should.

Why do you say that custom is king over all? To me it seems to be so; for I have myself had intercourse with kings, and among others also Croesus the Lydian, who was greatly superior to other kings both in happiness of fortune and also in magnificence of wealth: now this man, when he was at his height of prosperity, was overcome by custom; and being overcome by it he became a slave instead of remaining free.

How do the Babylonians give their daughters in marriage? The women among the Babylonians are wont to marry as follows: when a girl has come to be of age for marriage, her parents summon an assembly of all who dwell about them and bid them choose out each one of them which man he most likes; and when they have gone through them all and declared their opinions, it is necessary that the parents do whatsoever shall have been voted by the greatest number.

Tell me of the gold-digging ants of India, larger than foxes. O ye men of Hellas! I know that there are a race called Ants, and their size is great; but whether they be so large as to have been called from this name by the Indians, namely "ants," that I am not able to say. For I heard from them nothing but marvels besides all these which they told about the gold-digging ants.

Tell me of the Egyptians and how their customs differ from those of other men. Stranger to Hellas, this race has been by nature most prone to novelty of any nation that I know; for when they have heard but a word or seen it but once, straightway they forthwith are acquainted with all. Whenever any new thing comes in from abroad they are more eager than any other nation to learn about it; but when the thing has been learnt by them, forthwith they lay aside all eagerness to know and do not care at all.

Running it

Q5_K_M GGUF, served with Ollama. Pair it with the elicitation frame (a "traveller asks Herodotus" template); see the Modelfile in the project repo. Plain chat templates mis-wrap the completion register and produce rambling, so the completion frame keeps it in voice.

Intended uses

The Elect console (Ollama serve plus chat)
Historical dialogue and character embodiment
Research into the Herodotean register: digression, wonder, ethnography, fate

Limitations

This is a 7B model, not a search index over The Histories. It invents freely: names, dates, numbers, sources, events. It imitates the manner of Herodotus, the long digression and the reported marvel, but the specific claims it makes are not reliable history. Treat every output as the voice, never as the record, and check anything before you repeat it.

License

CC-BY-NC-4.0. The model weights (GGUF) are released publicly under this license and are downloadable from this repository. The training corpus is not bundled: it is rebuildable from the cited public-domain source (Macaulay 1890, Project Gutenberg #2707 and #2456).