CypressTree-0.5-7B

CypressTree-0.5-7B

모델 설명

이 모델은 텍스트(자연어)를 SQL 쿼리로 변환하는 데 특화된 Mistral-7B-Instruct-v0.3 의 미세 조정(Fine-Tuning) 버전입니다.

• 모델명: CypressTree-0.5-7B
• 기반 모델: Mistral-7B-Instruct-v0.3
• 기반 모델 설명: Mistral-7B-v0.3을 명령어에 맞게 미세 조정(Instruct Fine-Tuned)한 대규모 언어 모델(LLM)입니다. Mistral-7B-v0.2 대비 확장된 32768 어휘, v3 토크나이저, 함수 호출 지원 등의 개선 사항을 포함합니다.
• 학습 데이터셋: gretelai/synthetic_text_to_sql
• 데이터셋 설명: Gretel Navigator를 활용하여 생성된 고품질의 text-to-sql 샘플로 구성된 풍부한 데이터셋이며, Apache 2.0 라이선스로 배포됩니다.

프롬프트 템플릿

<s>
### Instruction:
{question}

### Context:
{schema}

### Response:

---

사용법

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "namuai-x/CypressTree-0.5-7B"
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False)

text = """<s>
### Instruction:
What is the total volume of timber sold by each salesperson, sorted by salesperson?

### Context:
CREATE TABLE salesperson (salesperson_id INT, name TEXT, region TEXT); INSERT INTO salesperson (salesperson_id, name, region) VALUES (1, 'John Doe', 'North'), (2, 'Jane Smith', 'South'); CREATE TABLE timber_sales (sales_id INT, salesperson_id INT, volume REAL, sale_date DATE); INSERT INTO timber_sales (sales_id, salesperson_id, volume, sale_date) VALUES (1, 1, 120, '2021-01-01'), (2, 1, 150, '2021-02-01'), (3, 2, 180, '2021-01-01');

### Response:
"""

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)

inputs = tokenizer(text, return_tensors="pt")
inputs = {key: value.to(device) for key, value in inputs.items()}

outputs = model.generate(**inputs, max_new_tokens=300, pad_token_id=tokenizer.eos_token_id)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

---

출력 예시

### Instruction:
What is the total volume of timber sold by each salesperson, sorted by salesperson?

### Context:
CREATE TABLE salesperson (salesperson_id INT, name TEXT, region TEXT); INSERT INTO salesperson (salesperson_id, name, region) VALUES (1, 'John Doe', 'North'), (2, 'Jane Smith', 'South'); CREATE TABLE timber_sales (sales_id INT, salesperson_id INT, volume REAL, sale_date DATE); INSERT INTO timber_sales (sales_id, salesperson_id, volume, sale_date) VALUES (1, 1, 120, '2021-01-01'), (2, 1, 150, '2021-02-01'), (3, 2, 180, '2021-01-01');

### Response:
SELECT salesperson.name, SUM(timber_sales.volume) as total_volume FROM salesperson JOIN timber_sales ON salesperson.salesperson_id = timber_sales.salesperson_id GROUP BY salesperson.name ORDER BY total_volume DESC;

---

하드웨어 및 소프트웨어

• 학습 하드웨어: 4개의 Tesla V100-PCIE-32GB GPU

---

라이선스

• Apache-2.0

---

연락처

• [email protected]

---

CypressTree-0.5-7B

Model Description

This model is a fine-tuned version of Mistral-7B-Instruct-v0.3, specialized in converting natural language text into SQL queries.

• Model Name: CypressTree-0.5-7B
• Developed by: namuai-x
• Base Model: Mistral-7B-Instruct-v0.3
• Base Model Description: The Mistral-7B-Instruct-v0.3 Large Language Model (LLM) is an instruct fine-tuned version of Mistral-7B-v0.3. It includes improvements over Mistral-7B-v0.2, such as an extended vocabulary of 32768, v3 Tokenizer support, and function calling capabilities.
• Dataset Name: gretelai/synthetic_text_to_sql
• Dataset Description: gretelai/synthetic_text_to_sql is a rich dataset of high-quality synthetic Text-to-SQL samples, generated using Gretel Navigator and released under the Apache 2.0 license.

Prompt Template

<s>
### Instruction:
{question}

### Context:
{schema}

### Response:

---

How to Use it

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "namuai-x/CypressTree-0.5-7B"
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False)

text = """<s>
### Instruction:
What is the total volume of timber sold by each salesperson, sorted by salesperson?

### Context:
CREATE TABLE salesperson (salesperson_id INT, name TEXT, region TEXT); INSERT INTO salesperson (salesperson_id, name, region) VALUES (1, 'John Doe', 'North'), (2, 'Jane Smith', 'South'); CREATE TABLE timber_sales (sales_id INT, salesperson_id INT, volume REAL, sale_date DATE); INSERT INTO timber_sales (sales_id, salesperson_id, volume, sale_date) VALUES (1, 1, 120, '2021-01-01'), (2, 1, 150, '2021-02-01'), (3, 2, 180, '2021-01-01');

### Response:
"""

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)

inputs = tokenizer(text, return_tensors="pt")
inputs = {key: value.to(device) for key, value in inputs.items()}

outputs = model.generate(**inputs, max_new_tokens=300, pad_token_id=tokenizer.eos_token_id)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

---

Example Output

### Instruction:
What is the total volume of timber sold by each salesperson, sorted by salesperson?

### Context:
CREATE TABLE salesperson (salesperson_id INT, name TEXT, region TEXT); INSERT INTO salesperson (salesperson_id, name, region) VALUES (1, 'John Doe', 'North'), (2, 'Jane Smith', 'South'); CREATE TABLE timber_sales (sales_id INT, salesperson_id INT, volume REAL, sale_date DATE); INSERT INTO timber_sales (sales_id, salesperson_id, volume, sale_date) VALUES (1, 1, 120, '2021-01-01'), (2, 1, 150, '2021-02-01'), (3, 2, 180, '2021-01-01');

### Response:
SELECT salesperson.name, SUM(timber_sales.volume) as total_volume FROM salesperson JOIN timber_sales ON salesperson.salesperson_id = timber_sales.salesperson_id GROUP BY salesperson.name ORDER BY total_volume DESC;

---

Hardware and Software

• Training Hardware: 4 Tesla V100-PCIE-32GB GPUs

---

License

• Apache-2.0

---

Contact

• [email protected]