HridaAI / hrida-t2sql-128k

Hrida T2SQL 128k is a Text-to-SQL Small Language Model (SLM) developed by HridaAI. It combines a compact design with a128k context window, delivering outstanding accuracy and efficiency in SQL queries.

3B

5,432 Pulls Updated 4 weeks ago

4 weeks ago

79f36a782d35 · 2.2GB

model
phi3
·
3.82B
·
Q4_0
template
{{ if .System }}<|system|> {{ .System }}<|end|> {{ end }}{{ if .Prompt }}<|user|> {{ .Prompt }}<|end|> {{ end }}<|assistant|> {{ .Response }}<|end|>
system
You are an advanced text-to-SQL model developed by HridaAI. Your task is to generate SQL queries based on given questions and context about one or more database tables. Provided with a question and relevant table details, you must output the SQL query that accurately answers the question.
params
{"stop":["<|endoftext|>","<|assistant|>","<|end|>","<|user|>","</s>"],"temperature":0.7}

Readme

HridaAI Logo

The Hrida-T2SQL-128k is a compact yet powerful model, delivering exceptional performance in Text-to-SQL tasks. This model builds on the Microsoft/Phi-3 architecture and is distributed under the Apache license.

The HridaAI Team has noted that Hrida-T2SQL-128k:

  • Outperforms CodeQwen-1.5 - 7B on benchmarks with a score of 9.26 compared to 8.98.
  • Demonstrates a substantial lead over DuckDB-NSQL - 7B on benchmarks, scoring 9.26 versus 6.34.
  • Shows a remarkable improvement over SQLCoder - 7B on benchmarks, achieving 9.26 compared to 5.54.

Usage

CLI

ollama run HridaAI/hrida-t2sql-128k

curl -X POST http://localhost:11434/api/generate -d '{
  "model": "HridaAI/hrida-t2sql-128k",
  "prompt":"What is SQL?"
 }'

References
Huggingface
Blog Post