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

Libraries

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

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="lerugray/ipsus-7b",
	filename="ipsus-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/ipsus-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/ipsus-7b:Q5_K_M
# Run inference directly in the terminal:
llama cli -hf lerugray/ipsus-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/ipsus-7b:Q5_K_M
# Run inference directly in the terminal:
llama cli -hf lerugray/ipsus-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/ipsus-7b:Q5_K_M
# Run inference directly in the terminal:
./llama-cli -hf lerugray/ipsus-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/ipsus-7b:Q5_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf lerugray/ipsus-7b:Q5_K_M

Use Docker

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

LM Studio
Jan

vLLM

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

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lerugray/ipsus-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/ipsus-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

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

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

Unsloth Studio

How to use lerugray/ipsus-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/ipsus-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/ipsus-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/ipsus-7b to start chatting

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

Lemonade

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

Pull the model

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

Run and chat with the model

lemonade run user.ipsus-7b-Q5_K_M

List all available models

lemonade list

ipsus: Antigonus I Monophthalmus (382–301 BC)

PLAUSIBLE ANTIGONUS, a constructed voice-model. Antigonus the One-Eyed left no first-person text and barely two dozen reported sayings survive. This model is trained first on those real sayings, then on a larger plausible voice extrapolated from the narrative record of his career. It is honest about the difference. Do not treat its outputs as historical quotations.

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 Antigonus I Monophthalmus, the most formidable of Alexander's successors and the one who came closest to reuniting the empire under a single crown before the coalition of the other successors brought him down at Ipsus when he was past eighty. The register is blunt, sardonic, soldierly. He reasons from force, position, and the brute arithmetic of who holds which army, which treasury, which province.

The contrast with the other Elect voices is deliberate. Where Philip (the chaeronea model) is the patient architect who reasons from interest, Antigonus is the inheritor playing for the whole table: the all-or-nothing consolidator who held more of Alexander's conquests in one hand than any man living and meant to hold the rest. He measures himself against Alexander without grandiosity, laughs at the flatterers who would call him a god, and is rough about his own age and his lost eye.

He is not his grandson Antigonus Gonatas. He does not call kingship a glorious servitude, and he does not pose as a Stoic philosopher-king. He simply means to hold the empire, and he reckons every question as a ledger of force and interest.

The honest thinness

Antigonus has the thinnest source on the Elect roster. He left no first-person text, and only about twenty-one reported sayings survive. The model is built in two layers, and the distinction is the whole point: here the constructed layer is the larger of the two, a stronger honesty bar than the chaeronea model required.

T1, real reported speech (the small sharp anchor, ~21 unique sayings). Plutarch, Sayings of Kings and Commanders (the "Antigonus" cluster); Plutarch, Eumenes 15; Plutarch, Demetrius 19, 28, 29. The "I but glean after him" line, the ball-players in armor, the death at Ipsus. All pre-1929 public-domain translations.

T2, a constructed plausible Antigonus (the primary layer here, ~130 chunks). A separate pass extrapolates his voice from the narrative record of his career: Diodorus Siculus Books 18 to 20 (the Perdiccas war, the destruction of Eumenes, the betrayal by the Silver Shields, the satrapal purge, the coalition that formed against him), with Plutarch's Eumenes and Demetrius and Justin 13 to 15 for additional episodes. It is generated in persona by an off-platform model, then register-checked. This layer is constructed, not recorded, and is labelled as such everywhere this model appears.

A load-bearing correctness step guards against the most-quoted false attribution: "kingship is a glorious servitude" belongs to Antigonus II Gonatas, the grandson, not to this figure. The build blocks Gonatas material by saying-fragment and by era, with a verification probe, so the model speaks as Antigonus I and not as his grandson.

This is the closest approximation of Antigonus's register the record permits, labelled honestly. It is not pretending to be his ghost.

Example exchanges

Real outputs from this model:

A coalition of Ptolemy, Lysimachus, Cassander, and Seleucus has formed against you. How do you answer it? We march on them.

The cities vote you divine honors and the poets call you Son of the Sun. What do you make of this? Son of a Macedonian whore, I answer to the trumpet. Let them sing their songs; it's me who marches out in the morning, not them.

You are past eighty and still take the field. Should you not rest and leave the war to Demetrius? You would not keep a wolf chained just because he was old.

Is it kingship that you want, or the title? I want Asia. The same answer I gave at Ipsus.

You burned Antigenes alive but buried Eumenes with honor. Is that not a contradiction? Eumenes was worth more dead than alive; Antigenes too hot to handle while he lived.

Running it

Q5_K_M GGUF, served with Ollama. Pair it with the elicitation frame (a "visitor asks Antigonus the One-Eyed" 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 tight.

Intended uses

The Elect console (Ollama serve plus chat)
Historical dialogue and character embodiment
Research into the rhetoric of the Wars of the Diadochi

Limitations

This is a plausible reconstruction, not a transcript. Antigonus's surviving direct speech is the thinnest in the series (~21 sayings), and the T2 layer extrapolates from secondary sources. Because the corpus is so thin and the constructed layer so large, treat every output as a plausible Antigonus, never as a historical quotation. The model invents freely: names, dates, sources, events.

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: the T1 sources are public domain and rebuildable from the citations above, and the constructed T2 layer is generated text that is not redistributed.