vindows/qwen2.5-7b-text-to-sql-merged

Qwen2.5-7B Merged Model for Text-to-SQL

This is a fully merged model (base + LoRA) ready for direct use. No need to load adapters separately!

Quick Links

• 🔧 LoRA Adapter Only: vindows/qwen2.5-7b-text-to-sql
• 🖥️ GGUF for CPU: vindows/qwen2.5-7b-text-to-sql-gguf

Performance

Spider Benchmark (200 examples)

Metric	Score
Exact Match	0.00%
Normalized Match	0.50%
Component Accuracy	92.60%
Average Similarity	25.47%

Training Metrics

Metric	Base	Fine-tuned	Improvement
Loss	2.1301	0.4098	80.76% ⬆️
Perplexity	8.4155	1.5064	82.10% ⬆️

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained(
    "vindows/qwen2.5-7b-text-to-sql-merged",
    device_map="auto",
    torch_dtype=torch.bfloat16,
    trust_remote_code=True
)

tokenizer = AutoTokenizer.from_pretrained(
    "vindows/qwen2.5-7b-text-to-sql-merged",
    trust_remote_code=True
)

# Generate SQL from natural language
prompt = """Convert the following natural language question to SQL:

Database: concert_singer
Question: How many singers do we have?

SQL:"""

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=128, temperature=0.1, do_sample=False)
result = tokenizer.decode(outputs[0], skip_special_tokens=True)

# Extract SQL (remove prompt and extra text)
sql = result.split("SQL:")[-1].strip().split('\n\n')[0]
print(sql)

Model Size

• Parameters: 7B
• Disk Size: ~16GB
• Recommended GPU: 24GB+ VRAM

Limitations

See the LoRA adapter model card for detailed limitations and recommendations.

License

Apache 2.0