Instructions to use qingy2024/NaturalLM-7B-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use qingy2024/NaturalLM-7B-Instruct with Transformers:

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

pipe = pipeline("text-generation", model="qingy2024/NaturalLM-7B-Instruct")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("qingy2024/NaturalLM-7B-Instruct")
model = AutoModelForCausalLM.from_pretrained("qingy2024/NaturalLM-7B-Instruct")
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 qingy2024/NaturalLM-7B-Instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "qingy2024/NaturalLM-7B-Instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "qingy2024/NaturalLM-7B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/qingy2024/NaturalLM-7B-Instruct

SGLang

How to use qingy2024/NaturalLM-7B-Instruct 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 "qingy2024/NaturalLM-7B-Instruct" \
    --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": "qingy2024/NaturalLM-7B-Instruct",
		"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 "qingy2024/NaturalLM-7B-Instruct" \
        --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": "qingy2024/NaturalLM-7B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Studio new

How to use qingy2024/NaturalLM-7B-Instruct with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for qingy2024/NaturalLM-7B-Instruct to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for qingy2024/NaturalLM-7B-Instruct to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for qingy2024/NaturalLM-7B-Instruct to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="qingy2024/NaturalLM-7B-Instruct",
    max_seq_length=2048,
)

Docker Model Runner
How to use qingy2024/NaturalLM-7B-Instruct with Docker Model Runner:
```
docker model run hf.co/qingy2024/NaturalLM-7B-Instruct
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

This Mistral 7B fine-tune is trained (for 150 steps) to talk like a human, not a "helpful assistant"!

It's also very beta right now. The dataset (qingy2024/Natural-Text-ShareGPT) can definitely be improved.

Example with temperature = 0.8, repeat penality = 1.0, top k sampling = 40:

User: Write a short essay about why human cloning is unethical.

NaturalLM-7B-Instruct: I'd be happy to do that. I'll start by saying that I am not against all kinds of human cloning. There are many potential benefits of therapeutic cloning. For example, you could clone a patient's cells and grow them in culture so you could grow new organs from those cells for that patient. It could be very useful for a lot of diseases. But I think there are very clear ethical reasons why reproductive cloning should not be allowed. The first reason is that reproductive cloning would violate the basic right to a unique human identity. We all have unique fingerprints, we all have unique DNA, we all have unique histories. If we were to create a clone, we would create a person with no unique human identity. The second reason is that reproductive cloning would violate the basic human dignity of the cloned person. The cloned person would be treated as a mere means to an end. The cloning parents would view the cloned child as a replacement for them, as a second try for them, not as a unique human being in his or her own right. I think that that would be a great disrespect for the basic human dignity of the cloned person. A third reason is that reproductive cloning would create a very dangerous slippery slope. Once we cross the line into reproductive cloning, what's to stop us from creating a designer child? We could use reproductive cloning to select for things like intelligence, beauty, athletic prowess, even traits like blue eyes. But those are all traits that are part of what makes us human. They're part of our human diversity. We should not be creating human beings to fit a certain mold.

Downloads last month: 6

Safetensors

Model size

7B params

Tensor type

F16

Model tree for qingy2024/NaturalLM-7B-Instruct

Quantizations

7 models

Space using qingy2024/NaturalLM-7B-Instruct 1

Collection including qingy2024/NaturalLM-7B-Instruct

NaturalLM

Collection

A Natural-Sounding LLM • 5 items • Updated Mar 2 • 1