README.md

---
library_name: peft
base_model: Qwen/Qwen1.5-1.8B-Chat
datasets:
- b-mc2/sql-create-context
language:
- en
pipeline_tag: text-generation
---

# Qwen-1.5-1.8B-SQL Model

## Description
This model, `deltawi/Qwen-1.5-1.8B-SQL`, is fine-tuned on SQL generation based on questions and context. It's designed to generate SQL queries from natural language descriptions, leveraging the [Qwen 1.5 - 1.8B model](https://huggingface.co/Qwen/Qwen1.5-1.8B-Chat).

## Installation
To use this model, you need to install the `transformers` library from Hugging Face. You can do this using pip:
```bash
pip install transformers huggingface_hub accelerate peft
```

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

# Set the device
device = "cuda"  # replace with your device: "cpu", "cuda", "mps"

from transformers import AutoModelForCausalLM, AutoTokenizer
import random 


peft_model_id = "deltawi/Qwen-1.5-1.8B-SQL"
base_model_id = "Qwen/Qwen1.5-1.8B-Chat"

device = "cuda"

model = AutoModelForCausalLM.from_pretrained(base_model_id, device_map="auto")
model.load_adapter(peft_model_id)
tokenizer = AutoTokenizer.from_pretrained(
        "deltawi/Qwen-1.5-1.8B-SQL",
        #model_max_length=2048,
        padding_side="right",
        trust_remote_code=True,
        pad_token='<|endoftext|>'
    )

# Define your question and context
Question = "Your question here"
Context = """
Your SQL context here
"""

# Create the prompt
prompt = f"Question: {Question}\nContext: {Context}"
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": prompt}
]

# Prepare the input
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(device)

# Generate the response
generated_ids = model.generate(
    model_inputs.input_ids,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

# Decode the response
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

```

## More details

- Base Model: Qwen 1.5-1.8B
- Fine-tuned for: SQL Query Generation
- Fine-tuning using LoRA: r=64
- Training Data: [b-mc2/sql-create-context](https://huggingface.co/datasets/b-mc2/sql-create-context)

### Framework versions

- PEFT 0.8.2