README.md

metadata

tags:
  - gguf
  - llama.cpp
  - unsloth
license: mit
datasets:
  - b-mc2/sql-create-context
language:
  - en
metrics:
  - accuracy
base_model:
  - microsoft/Phi-3-mini-4k-instruct

This is a specialized Text-to-SQL model fine-tuned from the Microsoft Phi-3-mini-4k-instruct architecture. It has been optimized using Unsloth to provide high-accuracy SQL generation while remaining lightweight enough to run on consumer hardware.

Key Features

• Architecture: Phi-3-mini (3.8B parameters)
• Quantization: Q4_K_M GGUF & Q5_K_M
• Training Technique: Fine-tuned using Lora with Unsloth.
• Format: GGUF (Ready for Ollama, LM Studio, and llama.cpp)
  • phi-3-mini-4k-instruct.Q4_K_M.gguf
  • phi-3-mini-4k-instruct.Q5_K_M.gguf

Usage Instructions

Ollama (Recommended)

To deploy locally:

1. Download the .gguf file.

2. Create the Modelfile with the following instructions

FROM ./phi-3-mini-4k-instruct.Q4_K_M.gguf

SYSTEM """You are a specialized SQL assistant. Your goal is to produce valid SQL queries based on the provided schema and question. Output only the SQL code and nothing else."""

TEMPLATE """<|system|>
{{ .System }}<|end|>
<|user|>
{{ .Prompt }}<|end|>
<|assistant|>
"""

PARAMETER stop "<|end|>"
PARAMETER temperature 0.1
PARAMETER num_ctx 2048
PARAMETER repeat_penalty 1.2

3. Run ollama create phi3-sql-expert -f Modelfile

4. Run ollama run phi3-sql-expert

Evaluation Data

The model was fine-tuned on the sql-create-context dataset, focusing on:

• Mapping natural language to complex SELECT, WHERE, and JOIN statements.
• Understanding table schemas provided in the prompt.
• Maintaining strict SQL syntax.

Recommended Settings

Temperature: 0.0 or 0.1 (SQL requires deterministic output).

Stop Tokens: Ensure <|end|> is set as a stop sequence to prevent "infinite looping" generation.

Context Window: 2048 tokens.

Model Developer: msquared

Base Model: Phi-3-mini-4k-instruct