MigueldsBatista/Arctic-Text2SQL-R1-7B-llamafile

Snowflake Arctic-Text2SQL-R1-7B - Llamafile

This repository provides a standalone, cross-platform executable llamafile for Snowflake/Arctic-Text2SQL-R1-7B.

What is a Llamafile?

A llamafile bundles the llamafile inference runtime together with model weights (Q4_K_M quantization) and default parameters into a single executable file. It runs locally on Linux, macOS, and Windows without needing Python, PyTorch, or package installations.

Quick Start

1. Download & Make Executable

curl -L -o arctic-text2sql-r1-7b.llamafile https://huggingface.co/MigueldsBatista/Arctic-Text2SQL-R1-7B-llamafile/resolve/main/arctic-text2sql-r1-7b.llamafile
chmod +x arctic-text2sql-r1-7b.llamafile

2. Run Interactive / CLI Mode

./arctic-text2sql-r1-7b.llamafile --cli -p "<|im_start|>user
Write a SQL query to list all department names.
<|im_end|>
<|im_start|>assistant
"

3. Run Server Mode (OpenAI-compatible API)

./arctic-text2sql-r1-7b.llamafile --server --port 8080

Access the API at http://localhost:8080/v1/chat/completions.

Hardware Acceleration Notes

• GPU Acceleration: Built-in CUDA acceleration works out-of-the-box on supported NVIDIA GPUs.
• CPU Fallback / Blackwell GPUs: For architectures not yet bundled in upstream llamafile (e.g. RTX 50-series sm_120) or systems without a dedicated GPU, append --gpu disable.