Use Docker images
docker run --gpus all \
--shm-size 32g \
-p 30000:30000 \
-v ~/.cache/huggingface:/root/.cache/huggingface \
--env "HF_TOKEN=<secret>" \
--ipc=host \
lmsysorg/sglang:latest \
python3 -m sglang.launch_server \
--model-path "Mahendra1742/SqlGPT" \
--host 0.0.0.0 \
--port 30000# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "Mahendra1742/SqlGPT",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'YAML Metadata Warning:The pipeline tag "text2text-generation" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, image-text-to-image, image-text-to-video, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other
🚀 SqlGPT LoRA fine-tuned on WikiSQL
This model is a LoRA fine-tuned version of Salesforce/codet5-small for natural language to SQL query generation on the WikiSQL dataset.
It uses PEFT (LoRA) to adapt the base model efficiently with minimal extra parameters.
Useful for learning and prototyping text-to-SQL tasks on simple table schemas.
📚 Training Details
- Base Model:
Salesforce/codet5-small - Adapter: LoRA (r=8, alpha=16) on attention
qandvmodules. - Dataset: WikiSQL (21k train, 3k val)
- Input Format:
question: <QUESTION> table: <TABLE_HEADERS> - Target: Human-readable SQL query
- Epochs: 1–3 recommended for small runs.
- Framework: 🤗 Transformers + PEFT
🧩 Example Usage
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
# Replace with your actual HF repo name
model_name = "Mahendra1742/SqlGPT"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
# Example input
question = "How many employees are in the Marketing department?"
table = "| department | employees |"
prompt = f"question: {question} table: {table}"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=128)
print("OUTPUT :- ")
print(" ")
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
Input
question: How many cities have a population over 1 million? table: | City | Population |
Output
SELECT COUNT(*) FROM table WHERE Population > 1000000
Install from pip and serve model
# Install SGLang from pip: pip install sglang# Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Mahendra1742/SqlGPT" \ --host 0.0.0.0 \ --port 30000# Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Mahendra1742/SqlGPT", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'