README.md

---
license: apache-2.0
language:
- en
library_name: transformers
pipeline_tag: text2text-generation
tags:
- code
- sql
- text-to-sql
- text2sql
- t2sql
base_model:
- microsoft/Phi-3.5-mini-instruct
---

Introducing Hrida-T2SQL-3B-V0.2, our latest small language model (SLM) tailored for data scientists and industry professionals. This advanced model marks a significant upgrade from our previous release, now equipped with 128k token context window as defaultfor handling even the most intricate data queries with precision. Powered by the Phi 3 architecture, it effortlessly converts natural language queries into precise SQL commands, enhancing data analysis efficiency and decision-making capabilities. 

Blog & benchmark coming soon!


## Prompt Template

```txt
### Instruction: 
Provide the system prompt.

### Dialect:
Specify the SQL dialect (e.g., MySQL, PostgreSQL, SQL Server, etc.).

### Context: 
Provide the database schema including table names, column names, and data types.

### Input: 
User's query.

### Response:
Expected SQL query output based on the input and context.

```

- **Instruction (System Prompt)**: This guides the model on processing input to generate the SQL query response effectively.
- **Dialect (Optional)**: Specify the SQL variant the model should use to ensure the generated query conforms to the correct syntax.
- **Context**: Provide the database schema to the model for generating accurate SQL queries.
- **Input**: Provide the user query for the model to comprehend and transform into an SQL query.
- **Response**: Expected output from the model.


## Chat Prompt Template

```txt
<s>
<|system|>
{ Instruction / System Prompt }
<|user|>
{ Context / User Query } <|end|>
<|assistant|>
```

## Run the Model

### Using Transformers

```python
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

# Define the model and tokenizer
model_id = "HridaAI/Hrida-T2SQL-3B-V0.2"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, trust_remote_code=True)

# Define the context and prompt
prompt = """
Answer to the query will be in the form of an SQL query.
### Context: CREATE TABLE Employees (
    EmployeeID INT PRIMARY KEY,
    FirstName VARCHAR(50),
    LastName VARCHAR(50),
    Age INT,
    DepartmentID INT,
    Salary DECIMAL(10, 2),
    DateHired DATE,
    Active BOOLEAN,
    FOREIGN KEY (DepartmentID) REFERENCES Departments(DepartmentID)
); 

CREATE TABLE Departments (
    DepartmentID INT PRIMARY KEY,
    DepartmentName VARCHAR(100),
    Location VARCHAR(100)
); 
### Input: if the hiring date of John is 25/12/2024 and he is 48 years old how long it will take him to retire lets say the retirement age is 55?
### Response:
"""
# Prepare the input
messages = [{"role": "user", "content": prompt}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)

# Generate the output
outputs = model.generate(inputs, max_length=300)
print(tokenizer.decode(outputs[0]))


```

### Using MLX

```python
from mlx_lm import generate, load

model,tokenizer = load("HridaAI/Hrida-T2SQL-3B-V0.2")

prompt = """
Answer to the quey will be in the form of SQL query.
### Context: CREATE TABLE Employees (
    EmployeeID INT PRIMARY KEY,
    FirstName VARCHAR(50),
    LastName VARCHAR(50),
    Age INT,
    DepartmentID INT,
    Salary DECIMAL(10, 2),
    DateHired DATE,
    Active BOOLEAN,
    FOREIGN KEY (DepartmentID) REFERENCES Departments(DepartmentID)
); 

CREATE TABLE Departments (
    DepartmentID INT PRIMARY KEY,
    DepartmentName VARCHAR(100),
    Location VARCHAR(100)
); ### Input: if the hiring date of John is 25/12/2024 and he is 48 years old how long it will take him to retire lets say the retirement age is 55?
### Response:"""

response = generate(model=model,tokenizer=tokenizer,prompt=prompt, verbose=True)

```