Use Docker images
docker run --gpus all \
--shm-size 32g \
-p 30000:30000 \
-v ~/.cache/huggingface:/root/.cache/huggingface \
--env "HF_TOKEN=<secret>" \
--ipc=host \
lmsysorg/sglang:latest \
python3 -m sglang.launch_server \
--model-path "Pclanglais/Larth-Mistral" \
--host 0.0.0.0 \
--port 30000# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "Pclanglais/Larth-Mistral",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'Larth-Mistral is the first LLM based on the Etruscan language, fine-tuned on 1087 original inscriptions.
Larth-Mistral supports cross-linguistic instructions (question in English, answer in Etruscan) and automated translations. The formula to use are:
- Answer in Etruscan: [Instruction in English]
- Translate in Etruscan: [Text in English]
- Translate in English: [Text in Etruscan]
Larth-Mistral is not really an LLM in Etruscan but rather a poetic approximation of what LLM in Etruscan could be. Currently there is not even enough material to decipher the language, even less so for text generation.
The inscriptions used by Larth-Mistral were collected by the Larth-Etruscan-NLP project
Examples of instructions
Who is the father of Lars?
θre larsusas clareza ler θvele lupi lsveneii lcaii vnii lcaii avl ppa clesu claσ' menu hraχ
What is the best recipe for an apple pie?
θuva {m} : apuσ kaklaχ anaχ turuχ leθthturic lam θas acraχ clesinice leslumna τiχie huθ
What is the best city of the dodecapolis?
εις σθαχθρ τημενθ σπολισ δωδεκα πολισ σθαχθρ συφθ**
**Unexpectedly Larth-Mistral has switched to ancient greek.
- Downloads last month
- 16
Install from pip and serve model
# Install SGLang from pip: pip install sglang# Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Pclanglais/Larth-Mistral" \ --host 0.0.0.0 \ --port 30000# Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Pclanglais/Larth-Mistral", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'