Instructions to use Vezora/Narwhal-7b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Vezora/Narwhal-7b with Transformers:

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

pipe = pipeline("text-generation", model="Vezora/Narwhal-7b")

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("Vezora/Narwhal-7b")
model = AutoModelForMultimodalLM.from_pretrained("Vezora/Narwhal-7b")

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Vezora/Narwhal-7b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Vezora/Narwhal-7b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Vezora/Narwhal-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Vezora/Narwhal-7b

SGLang

How to use Vezora/Narwhal-7b 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 "Vezora/Narwhal-7b" \
    --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": "Vezora/Narwhal-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "Vezora/Narwhal-7b" \
        --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": "Vezora/Narwhal-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Vezora/Narwhal-7b with Docker Model Runner:
```
docker model run hf.co/Vezora/Narwhal-7b
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Narwhal-7b

Model Created by Vezora

The Narwhal-7b is an innovative model comprising a blend of 60% Stable Beluga and 40% MegaCoder. This combination was further enhanced by integrating 40% Wizard-Math and 40% Llama Chat7b.

This synthesis has led to a model that demonstrates remarkable performance in mathematical tasks. It also maintains a robust ability to respond to various queries. During the testing phase, we employed both Stable Beluga and Llama Chat prompts, with the Llama v2 prompting yielding superior results. This improvement was likely a result of it being the final merge in the development process.

It's worth noting that the Narwhal-7b may stand as one of the best-performing models in its category. However, those interested in utilizing it must be aware of the commercial licenses associated with the underlying models. Due to some datasets used in training originating from OpenAI, this model is explicitly not intended for commercial use.

Benchmarks: Comprehensive benchmarking details will be available soon.

Downloads last month: 9

Model tree for Vezora/Narwhal-7b

Quantizations

1 model