Instructions to use defog/llama-3-sqlcoder-8b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use defog/llama-3-sqlcoder-8b with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="defog/llama-3-sqlcoder-8b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("defog/llama-3-sqlcoder-8b")
model = AutoModelForCausalLM.from_pretrained("defog/llama-3-sqlcoder-8b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use defog/llama-3-sqlcoder-8b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "defog/llama-3-sqlcoder-8b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "defog/llama-3-sqlcoder-8b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/defog/llama-3-sqlcoder-8b

SGLang

How to use defog/llama-3-sqlcoder-8b with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "defog/llama-3-sqlcoder-8b" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "defog/llama-3-sqlcoder-8b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "defog/llama-3-sqlcoder-8b" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "defog/llama-3-sqlcoder-8b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use defog/llama-3-sqlcoder-8b with Docker Model Runner:
```
docker model run hf.co/defog/llama-3-sqlcoder-8b
```

Any plans to release quantized model?

#15

by asusevski - opened Jul 12, 2024

Discussion

asusevski

Jul 12, 2024

I'd love to use a quantized version for local stuff, any chance we could get this released?

tensorsofthewall

Aug 3, 2024

•

edited Aug 3, 2024

You can try out this INT8 OpenVINO version here: https://huggingface.co/sandeshb/llama-3-sqlcoder-8b-int8-ov
Fair warning: I've played around with this quantized model for a couple of days, and it looks like there are some performance issues, particularly with negative answers (where answer should be "I don't know"). The model always returns an SQL query even if it isn't valid.

Using the default example on https://defog.ai/sqlcoder-demo, If I ask the question "What is the mass of the sun", then the full model returns "SELECT 'I do not know' AS answer;" whereas the quantized version returns a valid query as the answer, which will be incorrect for the question though i.e. "SELECT p.name, p.quantity FROM products p WHERE P.NAME = 'SUN'". However, I think the way to handle this is to just run the query and when the query output is null/zero, handle it accordingly. Otherwise, the quantized model works great.

asusevski

Aug 8, 2024

Oh thanks for pointing me in the right direction! At least verifying a bad sql query can be fast depending on the size of the database, so this might work well :)

asusevski changed discussion status to closed Aug 8, 2024

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment