Llama-3-Sqlcoder-8B-GGUF

SandLogic Technology - Quantized llama-3-sqlcoder-8b Models

Model Description

We have quantized the llama-3-sqlcoder-8b model into two variants:

1. Q5_KM
2. Q4_KM

These quantized models offer improved efficiency while maintaining performance.

Discover our full range of quantized language models by visiting our SandLogic Lexicon GitHub. To learn more about our company and services, check out our website at SandLogic.

Original Model Information

• Name: llama-3-sqlcoder-8b
• Developer: Defog, Inc.
• Model Type: Text-to-SQL generation
• Base Model: Meta-Llama-3-8B-Instruct
• Parameters: 8 billion
• License: CC-by-SA-4.0

Model Capabilities

The llama-3-sqlcoder-8b model is designed for generating SQL queries to answer questions, with support for Postgres, Redshift, and Snowflake databases. It has performance on-par with the most capable generalist frontier models.

Inference Parameters

• Temperature: 0 (no sampling)
• Prompt Format:

Generate a SQL query to answer this question: {user_question}
{instructions}
DDL statements:
{create_table_statements}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
The following SQL query best answers the question {user_question}:

Evaluation

The model was evaluated on SQL-Eval, a PostgreSQL-based evaluation framework developed by Defog for testing and alignment of model capabilities.

Intended Use Cases

1. SQL Generation: Automatically generate SQL queries based on natural language questions or instructions.
2. Database Interaction: Assist users in interacting with Postgres, Redshift, or Snowflake databases through text-based interfaces.
3. Data Analysis Support: Provide SQL-based solutions to data analysis problems described in natural language.
4. Programming Education: Help students learn SQL concepts and syntax by providing example queries and explanations.

Model Variants

We offer two quantized versions of the llama-3-sqlcoder-8b model:

1. Q5_KM: 5-bit quantization using the KM method
2. Q4_KM: 4-bit quantization using the KM method

These quantized models aim to reduce model size and improve inference speed while maintaining performance as close to the original model as possible.

Usage

pip install llama-cpp-python 

Please refer to the llama-cpp-python documentation to install with GPU support.

Basic Text Completion

Here's an example demonstrating how to use the high-level API for basic text completion:

from llama_cpp import Llama

llm = Llama(
    model_path="./model/llama-3-sqlcoder-8b.Q5_K_M.gguf",
    verbose=False,
    # n_gpu_layers=-1, # Uncomment to use GPU acceleration
    # n_ctx=2048, # Uncomment to increase the context window
)

output = llm.create_chat_completion(
    messages = [
        {"role": "system", "content": "You're an AI SQL coding  assistant who help in solving coding questions"},
        {
            "role": "user",
            "content": "write an simple sql table query and code to search employee name"
        }
    ]
)

print(output["choices"][0]['message']['content'])

Download

You can download Llama models in gguf format directly from Hugging Face using the from_pretrained method. This feature requires the huggingface-hub package.

To install it, run: pip install huggingface-hub

from llama_cpp import Llama

llm = Llama.from_pretrained(
    repo_id="SandLogicTechnologies/Llama-3-Sqlcoder-8B-GGUF",
    filename="*llama-3-sqlcoder-8b.Q5_K_M.gguf",
    verbose=False
)

By default, from_pretrained will download the model to the Hugging Face cache directory. You can manage installed model files using the huggingface-cli tool.

License

License: [CC-by-SA-4.0] Finetuned from model: [Meta-Llama-3-8B-Instruct]

Acknowledgements

We thank Defog, Inc. for developing the original llama-3-sqlcoder-8b model and the creators of Llama3 for their foundational work. Special thanks to Georgi Gerganov and the entire llama.cpp development team for their outstanding contributions.

Contact

For any inquiries or support, please contact us at [email protected] or visit our support page.