metadata

license: apache-2.0
base_model: codellama/CodeLlama-7b-Instruct-hf
tags:
  - text-to-sql
  - spider-dataset
  - sqlifyai
  - code-generation
library_name: transformers
pipeline_tag: text-generation

SQLifyAI - Text-to-SQL Model

This model was fine-tuned using SQLifyAI on the Spider dataset for converting natural language questions to SQL queries.

Model Details

Base Model: codellama/CodeLlama-7b-Instruct-hf
Dataset: Spider
Training: Multi-stage curriculum learning with advanced schema linking
Commit: 30-minute rapid test training run

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("dattheshshenoy/sqlifyai-30min-test")
model = AutoModelForCausalLM.from_pretrained("dattheshshenoy/sqlifyai-30min-test")

# Generate SQL
question = "What are the names of all students?"
schema = "CREATE TABLE students (id INT, name VARCHAR(50));"
prompt = f"### Question: {question}\n### Schema: {schema}\n### SQL:"

inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=128)
sql = tokenizer.decode(outputs[0], skip_special_tokens=True).split("### SQL:")[-1].strip()

Performance

Trained with advanced schema linking and curriculum learning
Optimized for Spider dataset evaluation metrics